A Protocol Framework for Adaptive Real-Time Communication in Industrial Wireless Sensor and Actuator Networks

Thesis for the degree of Doctor of Technology Sundsvall 2014

Wei Shen

Supervisors: Professor Tingting Zhang Professor Mikael Gidlund

Department of Information and Communication Systems Mid Sweden University, SE-851 70 Sundsvall, Sweden

ISSN 1652-893X Mid Sweden University Doctoral Thesis 178

ISBN 978-91-87557-31-6 Akademisk avhandling som med tillstand˚ av Mittuniversitetet i Sundsvall framlaggs¨ till offentlig granskning f or¨ avl aggande¨ av teknologie doktorsexamen, tisdagen den 06 mars 2014, klockan 13:15 i sal L111, Mittuniversitetet Sundsvall. Seminariet kommer att hallas˚ pa˚ engelska.

A Protocol Framework for Adaptive Real-Time Communication in Industrial Wireless Sensor and Actuator Networks

Wei Shen

c Wei Shen, 2014

Department of Information and Communication Systems Mid Sweden University, SE-851 70 Sundsvall, Sweden

Telephone: +46 (0)70 8569165

Printed by Kopieringen Mittuniversitetet, Sundsvall, Sweden, 2014 To My Wife iv Abstract

Low-power and resource-constrained wireless technology has been regarded as an emerging technology that introduces a paradigm shift in a wide range of applications such as industrial automation, smart grid, home automation and so on. The automation industry has significant contributions to economic revenues, job opportunities and world-class research. The low-power and resource-constrained wireless technology has brought new opportunity and challenges for industrial automation. The solutions of such wireless technology offer benefits in relation to lower cost and more flexible deployments/maintenances than the wired solutions, and new applications that are not possible with wired communication. However, these wireless solutions have been introducing new challenges. Wireless links are inherently unreliable, especially in industrial harsh environment, and wireless interference makes the problem even worse. Low-power consumption is required and real-time communication is generally crucial in industrial automation applications. This research work addresses that industrial wireless sensor and actuator network (IWSAN) should even be designed to provide service differentiation for wireless medium access and adapt to link dynamics for scheduling algorithms on top of real-time services. Specifically, exceeding the required delay bound for unpredictable and emergency traffic in industrial automation applications could lead to system instability, economic and material losses, system failure and, ultimately, a threat to human safety. Therefore, guaranteeing the timely delivery of the IWSAN critical traffic and its prioritization over regular traffic (e.g. non-critical monitoring traffic) is a significant topic. In addition, the state-of-the-art researches address a multitude of objectives for scheduling algorithms in IWSAN. However, the adaptation to the dynamics of a realistic wireless sensor network has not been investigated in a satisfactory manner. This is a key issue considering the challenges within industrial applications, given the time-constraints and harsh environments. In response to those challenges, a protocol framework for adaptive real-time communication in IWSAN is proposed. It mainly consists of a priority-based medium access protocol (MAC) and its extension for routing critical traffic, a hybrid scheme for acyclic traffic, and adaptive scheduling algorithms. To the best of our knowledge, the priority-based MAC solution is the first priority-enhanced MAC protocol compatible with industrial standards for IWSAN. The proposed solutions have been implemented in TinyOS and evaluated on a test-bed of Telosb motes and the TOSSIM network simulator. The experimental results indicate that the proposed priority-

v vi

based solutions are able to efficiently handle different traffic categories and achieve a significant improvement in the delivery latency. The hybrid scheme for acyclic traffic increases the throughput and reduces the delay compared to the current industrial standards. Numerical results show that the adaptive scheduling algorithms improve the quality of service for the entire network. They achieve significant improvements for realistic dynamic wireless sensor networks when compared to existing scheduling algorithms with the aim to minimize latency for real-time communication. Sammanfattning

Tr adl˚ osa¨ teknologier med mycket lag˚ effektf orbrukning¨ och liten minneskapacitet har blivit populara¨ sista decenniet och manga˚ nya applikationer inom olika domaner¨ som industriell automation, smarta elnat,¨ och hemautomation har m ojliggjorts.¨ Au- tomationsindustrin har lange¨ bidragit till ekonomisk tillvaxt,¨ nya arbetstillfallen¨ samt intressanta och nya sp annande¨ forskningsproblem har introducerats. Inom industriell automation sa˚ erbjuder tr adl˚ os¨ kommunikation en del f ordelar¨ j amf¨ ort¨ med tr adbunden˚ kommunikation genom st orre¨ flexibilitet, l agre¨ installationskostnader samt mojligheter¨ till nya applikationer som tidigare inte var m ojligt¨ med tr adbunden˚ kommunikation. Tr adl˚ os¨ kommunikation inneb ar¨ aven¨ en utmaning eftersom ra- diomiljon¨ inte ar¨ tillf orlitlig¨ i samma utstr ackning¨ som i tr adbunden˚ kommunikation, speciellt i industriella milj oer¨ dar¨ det kan vara mycket smuts, fukt, och interfer- ens fr an˚ andra tr adl˚ osa¨ sensorer och system. Detta forskningsarbete fokuserar pa˚ industriella tr adl˚ osa¨ sensornatverk¨ och hur man pa˚ b asta¨ satt¨ ska designa tr adl˚ osa¨ sensornatverk¨ for¨ bra tr adl˚ os¨ access i kraftigt varierande radiomiljo¨ samt algoritmer f or¨ effektiv schemal aggning¨ av real-tids applikationer. Inom industriell automation ar¨ det viktigt att realtidskraven p a˚ datakom- munikation halls˚ for¨ att inte skapa problem i produktionsprocessen som kan leda till instabilitet i systemen, oplanerade produktionsstopp som skapar ekonomisk f orlust¨ och i absolut v arsta¨ fall skador pa˚ m anniskor.¨ D arf¨ or¨ ar¨ det av yttersta vikt att man kan garantera att datatrafiken over¨ de tr adl˚ osa¨ sensornatverken¨ haller˚ de tid- skonstanter som ar¨ givna och att viktig datatrafik prioriteras f ore¨ data f or¨ mindre kritiska applikationer som t.ex monitorering. Vidare i detta forskningsarbete s a˚ undersoks¨ schemalaggningsalgoritmer¨ for¨ industriella tr adl˚ osa¨ sensornatverk¨ som hanterar varierande radiomiljoer.¨ I denna avhandling presenteras ett ramverk f or¨ protokoll passande till industriella tr adl˚ osa¨ sensornatverk.¨ Fokus ligger pa˚ adaptiv schemal aggning¨ och prior- itetsbaserad accessprotokoll. De f oreslagna¨ l osningarna¨ har implementerats i TinyOS och sedan utvarderats¨ i en testb add¨ bestaende˚ av TelosB sensornoder och i TOSSIM natverkssimulator.¨ Numeriska resultat visar att adaptiv schemal aggning¨ forb¨ attrar¨ kvaliteten i hela n atverket¨ jamf¨ ort¨ med existerande algoritmer for¨ schemal aggning.¨ Speciellt visar resultaten p a˚ att man kan minska f ordr¨ ojningen¨ i de tr adl˚ osa¨ sensornatverken¨ med adaptiv schemalaggning.¨ Vidare s a˚ visar experimentella resultat att det foreslagna¨ prioritetsbaserade accessprotokollet p a˚ ett effektivt s att¨ hanterar datatrafik med olika prioritet pa˚ ett f ored¨ omligt¨ satt.¨

vii viii Acknowledgements

First of all, I would like to thank my supervisor Prof. Tingting Zhang for her great supervisions during my PhD study. She has guided me past many obstacles in theoretical work. Thanks to Tingting for all the guidance, suggestions and ideas on my research work during these years, and for always believing me and supporting me. Without her patience and encouragement, this thesis would not have been possible. Thanks to my co-supervisor Prof. Mikael Gidlund for all his help and guidance in my research work. He has always given me supports and encouragements what I need. I have been taught by him that thinking the academic issues from the industrial application point of views. Thanks to Prof. Youzhi Xu for introducing me to this opportunity to start my Ph.D. study at Mid Sweden University. Thanks to Filip Barac and Felix Dobslaw for the excellent cooperations and friendliness during my research work. Thanks to Prof. M arten˚ Sj ostr¨ om¨ for his suggestions to improve the thesis writing. I would also like to thank Victor Kardeby, Stefan Forsstr om,¨ Patrik Osterberg,¨ Ulf Jennehag, Stefan Petterson, Annika Berggren and all other colleagues at the De- partment of Information and Communication Systems, Mid Sweden University for their kindness and friendliness. Last but not least, thanks to my families for their endless love and support. Spe- cial thanks to my dear wife, Jinlan Gao, for standing behind me in all that I do. This research work is ﬁnancially supported by KKS COINS project (20100258), Sensible Things That Communicate (STC) and Mid Sweden University.

Sundsvall, February 2014

Wei Shen

ix x Contents

Abstract v

Sammanfattning vii

Acknowledgements ix

List of Papers xiii

Terminology xix

1 Introduction 1 1.1 Overview ...... 1 1.2 Problem Description ...... 3 1.3 Objective and Research Topics ...... 4 1.4 Research Methodology ...... 5 1.5 Contributions ...... 6 1.6 Thesis Outline ...... 7

2 Background and Related Work 9 2.1 Low-Power and Resource-Constrained Wireless Technology ...... 9 2.2 Industrial Wireless Sensor and Actuator Networks ...... 11 2.3 Standardization Evolution of IWSAN ...... 14 2.4 MAC Protocols for WSN ...... 17 2.4.1 Fundamentals of Wireless MAC Protocols ...... 17 2.4.2 CSMA/CA Protocols ...... 21 2.4.3 MAC Protocols in Industrial Standards ...... 23

xi xii CONTENTS

2.4.4 Open Issues ...... 25 2.5 TDMA Scheduling ...... 26 2.5.1 Scheduling Problems ...... 26 2.5.2 Network Model and Wireless Conﬂict Model ...... 27 2.5.3 Scheduling Algorithms ...... 28 2.5.4 Open Issues ...... 30 2.6 Routing Protocols in WSN ...... 31 2.6.1 Collection Tree Protocol ...... 33 2.7 Implementation of Communication Protocols in WSN ...... 35

3 Summary of Contributed Papers 37

4 Conclusions and Future Work 45 4.1 Conclusions ...... 45 4.2 Future work ...... 46

Bibliography 47

PAPER I 57

PAPER II 85

PAPER III 117

PAPER IV 147

PAPER V 161

Biography 177 List of Papers

This thesis is mainly based on the following papers, herein referred by their Roman numerals:

Paper I PriorityMAC: A Priority-Enhanced MAC Protocol for Critical Trafﬁc in Industrial Wireless Sensor and Actuator Networks W. Shen , T. Zhang, F. Barac, and M. Gidlund Journal article, IEEE Transactions on Industrial Informatics , Vol. 10, No. 1, pp. 824-835, Feb. 2014.

Paper II CCA-Embedded TDMA Enabling Acyclic Trafﬁc in Industrial Wireless Sensor Networks D. Yang, M. Gidlund, W. Shen , Y. Xu, T. Zhang, and H. Zhang Journal article, Ad Hoc Networks , Vol. 11, Issue 3, pp. 1105-1121, May 2013.

Paper III SAS-TDMA: A Source Aware Scheduling Algorithm for Real-Time Com- munication in Industrial Wireless Sensor Networks W. Shen , T. Zhang, M. Gidlund, and F. Dobslaw Journal article, Wireless Networks , Vol. 19, No. 6, pp. 1155-1170, Aug. 2013. Paper IV Distributed Data Gathering Scheduling Protocol for Wireless Sensor Actor and Actuator Networks W. Shen , T. Zhang, and M. Gidlund Proceeding of Communications (ICC), 2012 IEEE International Conference on , pp. 7120-7125, Ottawa, Canada, 10-15 June 2012.

Paper V Joint Routing and MAC for Critical Trafﬁc in Industrial Wireless Sensor and Actuator Networks W. Shen , T. Zhang, and M. Gidlund Proceeding of Industrial Electronics (ISIE), 2013 IEEE International Symposium on , pp.1-6, 28-31 May 2013.

xiii xiv CONTENTS

Related articles, but not included in the thesis:

Paper VI Smart Border Routers for eHealthCare Wireless Sensor Networks W. Shen , Y. Xu, D. Xie, T. Zhang, and A. Johansson. Proceeding of 7th International Conference on Wireless Communications, Net- working and Mobile Computing (WiCOM) , pp.1-4, 23-25 Sept. 2011. Paper VII An ANT Based Wireless Body Sensor Biofeedback Network for Med- ical E-Health Care A. Johansson, W. Shen , and Y. Xu. Proceeding of 7th International Conference on Wireless Communications, Net- working and Mobile Computing (WiCOM) , pp.1-5, 23-25 Sept. 2011.

Paper VIII A Virtual Hypercube Routing Algorithm for Wireless Healthcare Net- works H. Huo, W. Shen , Y. Xu, and H. Zhang. Journal article, Chinese Journal of Electronics ,Vol.19, No.1, Jan. 2010. List of Figures

1.1 An example of future industrial automation networks [34] ...... 2 1.2 The protocol framework proposed in the thesis ...... 6

2.1 Standardization of low-power and resource-constrained wireless technologies ...... 10 2.2 A brief comparison of WirelessHART, ISA-100.11a, WIA-PA and IEEE 802.15.4e ...... 16 2.3 An example of the preamble sampling technique [47] ...... 19 2.4 Illustration of IEEE 802.15.4 superframe structure [2] ...... 23 2.5 Summary of WirelessHART packet format ...... 24 2.6 WirelessHART time slot and superframe [5] ...... 24 2.7 An example of cluster tree network architecture in IEEE 802.15.4 [5] . . 32 2.8 Illustration of the operation of CTP protocol ...... 34 2.9 A typical commercial development stack for WSNs ...... 35

3.1 PriorityMAC experiment shows it is able to efﬁciently handle different categories with different latency requirements...... 39 3.2 A comparison of nodes proportions with different PDR levels for SAS- TDMA algorithm and level-based algorithm with the refresh interval of 3.1s...... 41

xv xvi List of Tables

2.1 A comparison of low-power and resource-constrained wireless technologies ...... 11 2.2 Typical MCUs used in IEEE 802.15.4-compliant wireless devices . . . . 12 2.3 Typical transceivers used in IEEE 802.15.4-compliant wireless devices . 12 2.4 Notations of network model for scheduling algorithms ...... 28 2.5 Notations of network model for the level-based algorithm ...... 29

3.1 Objective, topics and my participation associated with corresponding papers ...... 38

xvii xviii Terminology

Abbreviations and Acronyms

WSN Wireless Sensor Networks WPAN Wireless Personal Area Network LR-WPAN Low Rate Wireless Personal Area Network SoC System on Chip IWSAN Industrial Wireless Sensor and Actuator Network MAC Medium Access Control DSSS Direct Sequence Spread Spectrum O-QPSK Offset Quadrature Phase-Shift Keying BPSK Binary Phase-Shift Keying PSSS Parallel Sequence Spread Spectrum ASK Amplitude Shift Keying FHSS Frequency Hopping Spread Spectrum GFSK Gaussian Frequency-Shift Keying OFDM Orthogonal Frequency-Division Multiplexing QAM Quadrature Amplitude Modulation PDR Packet Delivery Ratio nesC Network Embedded System C (programming language) MCU Microcontroller DCF Distributed Coordination Function DIFS Distributed Interframe Space CAP Contention Access Period CFP Contention Free Period GTS Guaranteed Time Slot AFS Adaptive Frequency Switch AFH Adaptive Frequency Hopping DSME Deterministic and Synchronous Multi-channel Extension GW Gateway CCA Clear Channel Assessment RTS Request To Send

xix xx LIST OF TABLES

CTS Clear To Send ETX Expected Transmission count RSS Received Signal Strength RAM Random Access Memory ROM Read Only Memory

Standards Organizations

IETF Internet Engineering Task Force IEC International Electrotechnical Commission IPSO Alliance Internet Protocol for Smart Objects Alliance ISA International Society of Automation ROLL Routing Over Low power and Lossy networks (IETF working group) CORE Constrained RESTful Environments (IETF working group) MANET Mobile Ad-hoc Networks (IETF working group)

Standards, Protocols and Algorithms

HART Highway Addressable Remote Transducer TDMA Time Division Multiple Access WirelessHART Wireless Highway Addressable Remote Transducer DBR Delay Bounded Routing protocol 6LoWPAN IPv6 over Low power WPAN RPL Routing Protocol for Low Power and Lossy Networks CoAP Constrained Application Protocol HTTP HyperText Transfer Protocol BT-LE Bluetooth Low Energy CSMA/CA Carrier Sense Multiple Access with Collision Avoidance WIA-PA Wireless network for Industrial Automation - Process Automa- tion TSMP Time Synchronized Mesh Protocol FDMA Frequency Division Multiple Access AES Advanced Encryption Standard TSCH Time-Synchronized Channel Hopping CDMA Code Division Multiple Access SDMA Space Division Multiple Access SMAC Sensor Medium Access Control protocol BMAC Berkeley Medium Access Control protocol WiseMAC Wireless Sensor Medium Access Control protocol UDP User Datagram Protocol LIST OF TABLES xxi

LEACH Low-Energy Adaptive Clustering Hierarchy protocol TEEN Threshold sensitive Energy Efﬁcient sensor Network protocol HEED Hybrid Energy-efﬁcient Distributed clustering protocol GBR Gradient Based Routing protocol GRAB GRAdient Broadcast protocol AODV Ad hoc On-Demand Distance Vector routing protocol OLSR Optimized Link State Routing protocol TBRPF Topology Dissemination Based on Reverse-Path Forwarding protocol DSR Dynamic Source Routing protocol CTP Collection Tree Protocol FTSP Flooding Time Synchronization Protocol PriorityMAC Priority-enhanced Medium Access Control protocol SAS-TDMA Source Aware Scheduling algorithm for TDMA based IWSAN xxii Chapter 1

Introduction

1.1 Overview

Low-power and resource-constrained wireless technology offering low data rate has been regarded as an emerging technology that introduces a paradigm shift in a wide range of applications such as industrial automation, smart grid, smart cities and urban networks, home automation, and structural health monitoring. Early con- ceptual designs of such technology can be traced back to the emergence of the distributed sensor networks program at the MIT Lincoln Labs in 1983 [1]. It aimed at developing and extending target surveillance and tracking technology in systems that employ multiple spatially distributed sensors and processing resources. In the following years, wireless sensor networks (WSN) have not only been demon- strated in research communities, but also attracted commercial interest. In order to facilitate interoperability that existing proprietary systems cannot provide, IEEE 802.15.4 Low Rate Wireless Personal Area Network (LR-WPAN) standard was initially released in 2003. The standard is a simple and low-cost communication network that allows wireless connectivity in applications with limited power and re- laxed throughput requirements [2]. It has been worldwide accepted and adopted far beyond its expected applications areas. Zigbee Alliance, Internet Engineering Task Force (IETF), International Electrotechnical Commission (IEC), Internet Pro- tocol for Smart Objects (IPSO) Alliance, Highway Addressable Remote Transducer (HART) Communication Foundation, International Society of Automation (ISA) and other international standards organizations have leveraged IEEE 802.15.4 as a base to build their upper standards. Multitude of leading semiconductors manufacturers worldwide made IEEE 802.15.4-compatible single-chip transceivers and/or system- on-chip (SoC) chips. Thousands of worldwide researchers have been attracted to conduct theoretical research, create prototypes and open pioneering companies. A growing number of global companies have been promoting IEEE 802.15.4-compatible products evolutions. Industrial automation is a wide area including process automation, factory au-

1 2 Introduction

Figure 1.1: An example of future industrial automation networks [34]

tomation, substation automation, building automation [3]. Obviously, automation industry has been significantly contributing to economic revenues, job opportunities and world-class research. According to a report from a number of Swedish automation companies and organizations in 2012, 60 percent of Swedish national household’s total export earnings comes from the process industry, which represents around 13% of Sweden’s total GDP [4]. The process automation will be regarded as the main case study to discuss in the thesis. Unless otherwise stated, “industrial automation” refers to “industrial process automation” in the rest of the thesis. Low-power and resource-constrained wireless technology has been bringing new opportunity and challenges for industrial automation, especially for communication technology in field networks. In a field network, industrial control and instrumenta- tion devices (such as sensors, actuators) provide process applications such as monitoring, measurement, alarms, diagnostics and process control [5–8] (an example of future industrial automation networks is shown in Fig. 1.1). Then, the concept, In- dustrial Wireless Sensor and Actuators Networks (IWSAN), is formed to represent the emerging technology. These applications typically have low-data-rate transmissions and small-size packets. Low-power consumption is required and real-time communication is generally crucial [8]. Industrial automation networks including field networks traditionally utilize wired communication such as HART protocol. HART communication foundation addresses that there are approximately 30 mil- lion HART devices installed and in service worldwide. The solutions of low-power and resource-constrained wireless technology offer benefits in relation to lower cost, 1.2 Problem Description 3

easier deployments and maintenance than the wired solutions. Therefore, a notable trend in industrial automation in recent years has been the replacement of wired communication with low power and low cost wireless sensor and actuator networks. However, such wireless solutions have been introducing new challenges. Wire- less links are inherently unreliable, especially in industrial harsh environments, and wireless interference makes the problem even worse. How to guarantee timely, reliable and secure communication becomes more signiﬁcant research topics. This research work addresses two vital issues on top of real-time service discussed in the following section. A series of novel solutions have been proposed, analyzed and evaluated after exploring related academic work and industrial standards.

1.2 Problem Description

Most of the existing WSN protocols primarily focusing on energy efficiency are unable to meet the real-time requirement in industrial automation applications. That is because most of them adopt contention-based or similar Medium Access Con- trol (MAC) protocols, which cannot provide deterministic behavior. But real-time communications require timely, reliable and predictable transmissions [9]. There- fore, standards in industrial automation, WirelessHART, ISA100.11a and WIA-PA, IEEE 802.15.4e which leverage IEEE 802.15.4 physical layer, introduce scheduling- based MAC protocols to provide a guaranteed access or bounded latency. The standards also adopt packet-level channel hopping to improve the reliability of individual wireless links by combating external interference and multi-path fading. This research work addresses that IWSAN should also be designed to provide and enhance service differentiation for wireless medium access and adapt to link dynamics for scheduling algorithms on top of real-time service. Firstly, a typical wireless industrial automation network deployment consists of field devices equipped with sensors and actuators that communicate with the plant network. Three categories of data traffic in industrial automation applications are: safety, control and monitoring [6]. The safety category refers to emergency action traffic, which is the most critical traffic class. The closed loop regulatory control, closed loop supervisory control and open loop control classes constitute the control category. The monitoring category consists of alerting class, in addition to the log- ging and downloading/uploading classes. The current standards mentioned above are able to provide guaranteed access (with a suitable scheduling algorithm) for periodic traffic such as monitoring. However, they are unable to provide lower latency in relation to unpredictable and emergency traffic because it is impossible for this type of traffic to have a priori assigned dedicated transmission time due to its non- deterministic occurrence. Exceeding a required delay bound for this type of traffic (e.g. an emergency safety action, an unpredictable critical control or an acyclic retransmission of critical control data) could lead to system instability, incur financial losses and even pose a threat to human safety [10]. Therefore, granting the prioritization of this critical traffic over regular traffic (e.g. non-critical and periodic traffic) is a fundamental problem. This is challenging for wireless technologies because the 4 Introduction

unpredictability regarding the arrival of the aperiodic traffic cause difficulties with regarding to making a suitable scheduling or reserving wireless bandwidth in order to guarantee its deadline. In order to solve this problem, the guaranteed access for non-critical traffic can be deferred or even destroyed. Thus, the real-time communication mechanisms with service differentiation for priority of critical traffic over non critical traffic is vital in IWSAN. Secondly, the requirement of dynamic is illustrated as follows. As mentioned, Time Division Multiple Access (TDMA), which is a typical scheduling-based protocol, is utilized in IWSAN standards. TDMA scheduling algorithms are not specified in the standards. Previous work on TDMA scheduling for WSN addressed several objectives, such as the determination of shortest schedules, or the finding of tradeoffs between delay and energy consumption. Nevertheless, the majority of existing approaches do not consider the effects of unreliable wireless links on scheduling algorithms. Link failures can result in schedules of broken order. Individual wireless links are inherently unreliable. Observations on real-platform testbeds with low- power wireless sensor networks show that links have wide ranges of packet delivery ratio (PDR) which can vary significantly over time, even without intra-collisions or high portions of external interference [11–13]. The dynamics of individual links are even more unpredictable in industrial environments, because of the human blockage, machinery and products obstacles. A centralized scheduling mechanism for industrial applications is thus taken as an example to describe this issue. In centralized scheduling, schedule requests come with topology and routing information when sent to the sink. The sink then runs a centralized scheduling algorithm and distributes the schedules to the sensors. Neighbors and routing information may change very frequently because of the unreliable nature of the individual wireless links and obstacles, thus generating heavy configurational overhead. This overhead in return increases the end-to-end delay. Therefore, IWSAN require scheduling algorithms which can adapt to link dynamics in real environment.

1.3 Objective and Research Topics

The overall aim of the research project is to propose solutions to solve vital challenges (such as real-time, service differentiation, adaption to link dynamics, security, wireless coexistence) in terms of applying emerging low-power and resource- constrained wireless technology to industrial automation applications. Objective: The objective of this research work is to design algorithms and protocols on top of real-time communications in IWSAN to:

• enhance service differentiation for TDMA-based wireless medium access

• adapt to link dynamics for scheduling algorithms

and to develop a prototype to evaluate and demonstrate potential of proposed solutions. 1.4 Research Methodology 5

Research topics: Speciﬁcally, the research topics are formulated into the following questions.

• (Paper I and V) Why is it a vital issue and challenge to provide service differentiation for TDMA-based wireless medium access? How to enable unpredictable critical trafﬁc to timely delivery and to have priority over real-time guaranteed non- or less-critical trafﬁc?

• (Paper II) How to improve the transmission efﬁciency for burst acyclic trafﬁc in WirelessHART and ISA100.11a based IWSAN?

• (Paper III) What are limitations of state-of-the-art scheduling algorithms in time division multiple access based wireless sensor networks in terms of industrial automation applications? How to adapt to link dynamics for scheduling algorithms in real environment?

• (Paper IV) Applications in industrial automation typically have low-data-rate transmissions and small-size packets, thus, there is typically a short-length sensor data with a relatively large overhead in each packet. How to efficiently schedule the transmissions in real environments to reduce traffic load, which, in turn, is useful for cost efficiency?

1.4 Research Methodology

The research methodologies used throughout the research work will be discussed in this section.

• Motivated by the realistic issues and challenges, exploring state-of-the-art work has been conducted. Then, a series of novel algorithms and protocols have been proposed after establishing a specific research objective. A layered architecture has been used to design protocols. Each layer has its own protocols and offers services to the layer above. The protocols of the various layers are called the protocol stack. The seven-layer OSI model is a well-known reference model for protocol stack design. This thesis work adopts five layers of the model: physical layer, data ink layer, network layer, transport layer and application layer, since the session layer and presentation layer are normally not specified in wireless sensor networks.

• The analytical methods using probability and random processes have been used to analyze the performance of proposed algorithms and protocols from theoretical standpoint. The performance such as delay has been formulated into mathematical equations to provide an intuition on the high level of the proposed algorithms and protocols. The theoretical results have been further compared with simulation results to evaluate to what extent they match each other. Besides, the equations can also be useful for the design purposes for real systems by deriving corresponding parameters in practice. 6 Introduction

• An implementation and the empirical evaluation on the implementation in a testbed is one of the most important methods in WSN research community. In order to develop a prototype to evaluate and demonstrate potential of proposed solutions, I have used a pragmatic approach to realize proposed solutions in a hardware testbed consisting of about 100 battery-powered resource- constrained embedded wireless devices. I have programmed using TinyOS [14] based on open sources such as TinyOS community [15], Contiki community [16], Berkeley OpenWSN [17].

• Most of the proposed solutions have been implemented in a simulator, TOSSIM [18], and some of them have been implemented in Matlab. The Matlab simulation has similar assumptions to ones in theoretical analysis so that it can be used to evaluated the correctness of theoretical analysis and vice versa. TOSSIM emulates the behavior of underlying raw hardware and it captures real-world environmental noise. Besides, TinyOS and TOSSIM is a widely accepted and used simulator in academic research community of wireless sensor network so that researchers can make fair comparisons between different solutions.

• The statistical methods (e.g., mean squared errors) have been utilized to evaluate differences between mathematical and simulation results; the experimental methods have been adopted to compare the proposed algorithms and protocols with state-of-the-art algorithms, protocols and standards.

1.5 Contributions

Figure 1.2: The protocol framework proposed in the thesis 1.6 Thesis Outline 7

The thesis proposes a protocol framework for adaptive real-time communication in industrial wireless sensor and actuator networks to achieve the research objective. Fig. 1.2 illustrates this protocol framework and speciﬁc components associated with contributed papers. Speciﬁcally speaking, the protocol framework consists of two MAC protocols presented in Paper I and II, two cross-layer scheduling protocols in Paper III and IV, and a routing protocol in Paper V.

• Paper I addresses a vital issue of IWSAN, namely, service differentiation for priority of critical traffic over non critical traffic, in real-time industrial automation applications. The paper proposes PriorityMAC, a Priority enhanced MAC protocol. To the best of our knowledge, this is the first priority-enhanced MAC protocol compatible with industrial standards for IWSAN. Paper V proposes Delay Bounded Routing (DBR) and extends the PriorityMAC to network layer by joint DBR and the PriorityMAC, designed for routing critical traffic in IWSAN.

• Paper II proposes a CCA-Embedded TDMA scheme that improves the transmission efficiency for burst acyclic traffic in WirelessHART and ISA-100.11a based industrial wireless networks. • Paper III presents SAS-TDMA, a Source-Aware Scheduling algorithm which uses a cross-layer solution to adapt to link dynamics in real environments. Pa- per IV further proposes a novel distributed scheduling protocol for data gathering transmission. It is also able to adapt the dynamics of links in a realistic low-power wireless network. In addition, the protocol reduces the scheduling traffic through maintaining distributed lightweight conflict-links tables. The purpose to adapt to link dynamics for scheduling algorithms significantly differentiates the contributions in the two papers from ones in most of state-of- the-art scheduling algorithms articles.

1.6 Thesis Outline

The remainder of the thesis is organized as follows. The background and related work about the topics covered in the thesis are discussed in Chapter 2. Chapter 3 presents summary of contributed papers. The conclusion and future work are provided in Chapter 4. 8 Chapter 2

Background and Related Work

In this chapter, a comprehensive overview of low power and resource constrained wireless technology, IWSAN and the standardization evolution of IWSAN, MAC protocols, TDMA scheduling, routing protocols and implementation in WSN is given and discussed.

2.1 Low-Power and Resource-Constrained Wireless Technology

Low power and resource constrained wireless technology and the advanced sensors/actuators technologies have been regarded as an emerging technology (i.e., WSN) applying in a wide range of applications. In order to facilitate interoperability that existing proprietary systems cannot provide, IEEE 802.15.4 LR-WPAN standard was initially released in 2003 [2]. As mentioned in Chapter 1, it has been worldwide accepted and applied far beyond its expected applications areas. Zigbee Alliance, IETF, IEC, IPSO Alliance, HART Communication Foundation, ISA and other international standards organizations have leveraged IEEE 802.15.4 as a base to build their upper standards. Multitude of leading semiconductors manufacturers worldwide made IEEE 802.15.4-compatible single-chip transceivers and/or SoC chips. Thou- sands of worldwide researchers have been attracted to conduct theoretical research, create prototypes and open pioneering companies. IEEE 802.15.4 becomes a de-facto standard in WSN. In 2012, the IEEE 802.15.4e standard [19] released with deﬁning a MAC amendment to the existing standard IEEE 802.15.4-2006 to better support industrial markets. Besides, various IETF working groups, IPv6 over Low power WPAN (6LoW- PAN), Routing Over Low power and Lossy networks (ROLL) and IETF Constrained RESTful Environments (CORE), have accomplished the integration of IEEE 802.15.4- compliant networks into the Internet, as shown in Fig. 2.1. In the network layer, the IETF 6LoWPAN working group has started in 2007 to specify a protocol, 6LoW-

9 10 Background and Related Work

Figure 2.1: Standardization of low-power and resource-constrained wireless technologies

PAN [20], for transmitting IPv6 over IEEE 802.15.4 networks by introducing 6LoW- PAN adaptation layer. On top of that, the IETF ROLL working group has standardized a Routing Protocol for Low Power and Lossy Networks (RPL) [21]. In application layer, a Constrained Application Protocol (CoAP) has been deﬁned by the IETF CORE working group with the purpose of running RESTful architectures such as the client/server model deﬁned by HyperText Transfer Protocol (HTTP) over low power and constrained wireless networks, meeting very low overhead and simplic- ity for constrained environments [22]. Recently, low-energy Bluetooth and low-power WiFi [23] become realistic. Fig. 2.1 also shows associated standardizations. Bluetooth Low Energy (BT-LE) [24], included in Bluetooth 4.0 released by Bluetooth Special Interest Group in 2010, is a radio technology targeted for devices that operate with coin cell batteries, which means that low power consumption is essential. BT-LE can also be integrated into existing Bluetooth devices so that devices such as mobile phones and computers can operate with existing Bluetooth accessories as well as BT-LE accessories. IEEE 802.11ah, targeting applications of wireless sensor networks, backhaul networks for sensor and meter and extended range of Wi-Fi, is being designed for supporting applications with the following requirements [25]: up to 8191 devices associated to an access point, adoption of power saving strategies, minimum network data rate of 100 kbps, operating carrier frequencies around 900 MHz, coverage up to 1 km in outdoor areas, one-hop network topology and short infrequent data transmissions. 2.2 Industrial Wireless Sensor and Actuator Networks 11

The standardization shall be completed approximately by March 2016 [26]. Table 2.1 shows a comparison of low-power and resource-constrained wireless technologies above regarding to radio modulation, transmission range, data rate, power saving mechanism, MAC and supported topology. As can be seen, the main features of low power and resource constrained wireless technologies include low data rate, low complexity of physical and MAC layer, low power consumption and etc. In order to further understand theses feature, typical Microcontrollers (MCUs) and transceivers used in IEEE 802.15.4-compliant wireless devices have been shown in Tables 2.2 and 2.3. We can see that theses MCUs have constrained computational capability with small size of memory and support ultra low power operating mode. Table 2.3 also shows that radios draw about the same current in transmit and receive mode. Improving the efﬁciency of power consumption for a system therefore consists in lowering the duty cycle of the radio as a whole, i.e. having the radio off most of the time [1].

Table 2.1: A comparison of low-power and resource-constrained wireless technologies

IEEE 802.15.4 Bluetooth-LE IEEE 802.11ah

Sub-1 GHz OFDM BPSK, 2.4 GHz DSSS O-QPSK 2.4 GHz FHSS GFSK QPSK, Modulation 868/915 MHz DSSS O-QPSK/BPSK (lower complexity than 16/64/256-QAM 868/915 MHz PSSS BPSK and ASK classic Bluetooth) (Ten times down clocking of 802.11ac)

Range up to 100 m up to 50 m up to 1 km

1 Mbps >100 Kbps Data rate up to 250 Kbps (up to 3Mbps in Classic Bluetooth) (one-tenth of 802.11ac)

Power saving Active/Inactive period Active/Low-power state Awake/Doze duty cycle

Enhanced Distributed CSMA/CA/ Coordination Function (EDCF) MAC FDMA/TDMA TDMA + channel hopping + contention-free access method

Topology Star, tree, mesh Star Star

2.2 Industrial Wireless Sensor and Actuator Networks

Recent years, wireless monitoring and control solutions in industrial automation have the beneﬁts of low cost and ﬂexible deployments and maintenance over wired solutions (e.g., Fieldbus system and wired HART) [27–31]. The low-power and resource-constrained wireless has been regarded as one of main appropriate 12 Background and Related Work

Table 2.2: Typical MCUs used in IEEE 802.15.4-compliant wireless devices

Manufacturer Size of RAM MCU core Other features and/or Module and Flash

STM32L: e.g., 16 KB RAM 1.65 V to 3.6 V power supply 32-bit ARM Cortex M3 STMicroelectronics’ 128 KB Flash 9 µA low power run mode STM32L series (typically) 0.9 µA standby mode < 8 µs wakeup time

10 KB RAM 1.8 V to 3.6 V power supply 16-bit MSP430 TI 128 KB Flash 2.1 µA standby mode (typically) 3.5 µs wakeup time

18 KB RAM 2.7 V to 3.3 V power supply 8-bit ATmega Atmel 128 KB Flash 8 mA in Active mode (typically) < 15 µA in Sleep mode

Table 2.3: Typical transceivers used in IEEE 802.15.4-compliant wireless devices

Current consumption Chips Manufacturer Supply voltage Sleep RX ON TX ON

CC2420 TI < 1 µA 18.8 mA 17.4 mA 2.1 V - 3.6 V

AT86RF231 Atmel 0.02 µA 12.3 mA 14.0 mA 1.8 V - 3.6 V

CC2530 (SoC, 0.4µA, 1µA, 24.0 mA 29.0 mA TI 2.0 V - 3.6 V MCU+Transceiver) 0.2 mA (MCU Idle) (MCU Idle) 2.2 Industrial Wireless Sensor and Actuator Networks 13

wireless technologies to apply in the industrial automation applications, especially for applications in field networks. In a field network, industrial control and instru- mentation devices (such as sensors, actuators) provide process applications such as monitoring, measurement, alarms, diagnostics and process control [5–8]. Then, the concept, IWSAN, is formed to represent the emerging technology. A notable trend in industrial automation in recent years has been the replacement of wired communication with IWSAN. The unique challenges and constraints of IWSANs have been attracting significant attentions in research community. The authors in [7,27,32–34] have investigated state-of-the-art and future requirements and challenges for IWSANs. Industrial automation networks traditionally utilize wired communication. Some requirements, such as real-time, safety, security, low-power consumption, backward compatibility, low-cost hardware, are essential for both wired and wireless solutions. But the major technical challenges between wired and wireless are different in order to meet those common requirements.

• Real-time: The real-time communications, which require timely, reliable and predictable transmissions [9], are essential for industrial automation applications.

• Safety: The automation devices should be designed to avoid or reduce the risk of uncontrolled or dangerous situations for safety of humans, environment and property [34].

• Security: The communication of IWSAN should be designed to avoid or reduce the denial-of-service attacks and intrusions [32].

• Low-power consumption: Wireless enables sensors and actuators to be ﬂexi- ble for deployment. IWSAN devices equipped with batteries may be deployed in bearings of motors, oil pumps, or hazardous environment [35] so as to be impractical to change batteries within a few years. Thus, low-power consump- tions of processor, transceivers, software communication stack are necessary and important.

• Backward compatibility: Unlike short-lifetime products in consumer market such as smart phones, devices in industrial automation normally have long lifetimes, e.g., 10 years [8]. Thus, it is important for IWSAN devices to have backward compatibility.

• Integration: The integration here refers to IWSAN should be efficiently and smoothly integrated into other existing infrastructures [34]. For example, a typical wireless field network consists of field devices normally communicate with the plant network.

• Low-cost hardware: The low-cost hardware is realized by the design of resource- constraint devices. 14 Background and Related Work

In summary, on the one hand, real-time and secure communication, design for safety are required for IWSAN. On the other hand, IWSAN has to meet the requirements of low-power consumption and low-cost hardware with limited computational ability. The major challenges come when attempting to meet the “conﬂicting” requirements. Furthermore, IWSAN has been introducing new challenges due to the inher- ent features of shared communication medium in wireless, self-organized and self- conﬁgured multi-hop networks and etc.

• Adapt to topology dynamics: As mentioned in Chapter 1, individual wireless links are inherently unreliable, which incurs topology. The dynamics of individual links are even more unpredictable in industrial environments, because of the human blockage, machinery and products obstacles. Thus, the adaption to topology changes is one of the challenges when designing communication stack for IWSAN. This is one of the two key issues addressed in this thesis. • Service differentiation for wireless medium access: This research work also addresses that IWSAN should also be designed to provide service differentiation for wireless medium access. The challenge is how to enable unpredictable critical trafﬁc to timely delivery and to have priority over real-time guaranteed non- or less-critical trafﬁc. • Wireless coexistence: IWSANs, such as WirelessHART and ISA-100.11a based wireless networks, share wireless medium with other wireless devices, which incurs the collisions of spectrum resource [36], i.e. coexistence problem. • Interoperability: The evolution of IWSAN in industrial automation applications is suffering from the interoperability problem among different industrial standards, such as WirelessHART, ISA-100.11a, WIA-PA, IEEE 802.15.4e and other standards. The new wireless technologies and resource-constrained IP based technologies may promote convergence solutions from a technique standpoint.

Industrial automation is a wide area and thus, there are wide variety of application requirements. Recent years, there are a series of industrial standards targeting at the process automation, especially for ﬁeld networks.

2.3 Standardization Evolution of IWSAN

The needs of low data rate and low power consumption for long battery lives and for large device amount have become signiﬁcant for sensors and control/actuators devices since more than ten years ago in wireless communication market. In order to facilitate interoperability that existing proprietary systems cannot provide, IEEE 802.15.4 standard was initially released in 2003, which becomes a de-facto standard in low-power and resource-constrained wireless sensor and actuator networks in the following years. 2.3 Standardization Evolution of IWSAN 15

ZigBee [37] is the most common standard among commercially available off-the- shelf solutions in the consumer market [38] and it is based on the IEEE 802.15.4 specification. It utilizes carrier sense multiple access with collision avoidance (CSMA/CA) as the multiple access control method, which can provide both a low delay and po- tentially sufficient throughput at low traffic loads without heavy external interference. This is particularly true for unpredictable traffic. Nevertheless, CSMA/CA is unable to provide guaranteed access to the wireless channel as the network density increases. It is also unable to serve the high number of devices within the specified cycle time in industrial applications [32]. Thus, CSMA/CA-based mechanisms are unsuitable for industrial automation applications, which have strict timing requirements [39, 40]. As a result, a series of industrial communication standards such as WirelessHART [5], ISA-100.11a [6], IEEE 802.15.4e [19] and WIA-PA [41] which have adopted IEEE 802.15.4 Physical layer have been released in recent years. TSMP A Time Synchronized Mesh Protocol (TSMP) which leverages IEEE 802.15.4 PHY was published in 2008. It is a medium access and networking protocol designed for the ratified WirelessHART standard in 2007 [42]. TSMP was standardized as one of the main proposed contributions in WirelessHART, ISA-100.11a and IEEE 802.15.4e. WirelessHART (IEC 62591) WirelessHART was the first open standard specifically designed for wireless communication in process measurement and control applications [43]. It officially presented by the HART Communication Foundation in September, 2007 aiming to be compatible with existing HART devices by adding wireless communication capability to the HART protocol. At the very bottom, it adopts IEEE 802.15.4-2006 as the physical layer. On top of that, WirelessHART defines its own time-synchronized MAC layer. The network layer supports self- organizing and self-healing mesh networking techniques. In this way, messages can be routed around interferences and obstacles [5, 43]. WirelessHART adopts a centralized routing scheme, in which a network manager is responsible for maintaining up-to-date routes and communication schedules for devices in the network. Wire- lessHART specification was approved by IEC as a full international standard (IEC 62591) in March 2010. A growing number of WirelessHART compatible products are available from major global suppliers such as ABB, Emerson, Endress+Hauser, Pepperl+Fuchs, Siemens. ISA-100.11a (IEC 62734) The ISA initiated work on a family of standards defin- ing wireless systems for industrial automation and control applications. The main purpose of the ISA100 committee is to provide a family of standards for industrial wireless networks, which will address the needs of the whole plant, such as process control, personnel and asset tracking and identification convergence of networks, and long-distance applications. ISA-100.11a is the first standard of the family and it was ratified in September 2009. It describes a mesh network designed to provide secure wireless communication to process control. ISA-100.11a received the IEC ap- proval as international standard (IEC 62734) in 2011. The global suppliers that make compatible products such as Honeywell, Yokogawa, Invensys, General Electric, Ma- soneilan, Yamataki, Fuji. 16 Background and Related Work

Figure 2.2: A brief comparison of WirelessHART, ISA-100.11a, WIA-PA and IEEE 802.15.4e

WIA-PA (IEC 62601) Chinese Industrial Wireless Alliance was established (in the leading of the Shenyang Institute of Automation, Chinese Academy of Science) in 2007 aiming to develop standards for industrial wireless network [44]. Wireless network for Industrial Automation - Process Automation (WIA-PA) is the first standard developed by the alliance for the urgent needs of process industries. WIA-PA defines the system architecture and communication specifications for the process automation. It was approved by IEC as international standard (IEC 62601) in 2011. IEEE 802.15.4e The IEEE802.15.4e standard released in 2012 [19] defines a MAC amendment to the existing standard IEEE 802.15.4-2006 to better support industrial markets. It announces to achieve better reliability and lower power consumption through time-synchronized channel hopping (TSCH). Fig. 2.2 presents a comparison of WirelessHART, ISA-100.11a, WIA-PA and IEEE 802.15.4e. As can be seen, they all build on top of IEEE 802.15.4 PHY with modi- fied MAC protocols. They have the same features in physical layer such as 250 kbps data rate, 16 channel direct sequence spread spectrum radio with 128-bit AES security encryption. In data link layer, TDMA and channel hopping are deployed in all of the standards with individual channel hopping mechanisms (TSCH in IEEE 802.15.4e is also TDMA based protocol). WirelessHART uses a slot time of 10 ms. 2.4 MAC Protocols for WSN 17

ISA-100.11a, WIA-PA and IEEE 802.15.4e uses a variable slot time, but defaults to 10 ms. In network layer, All of the standards speciﬁes or supports mesh network architecture with individual routing protocols to overcome obstacles, reach longer distances, create resilient paths for increasing reliability and etc. However, Wire- lessHART and WIA-PA do not support IPv6 packet format, while ISA-100.11a does by employing IETF 6LoWPAN protocol as an adaptation sublayer. WIA-PA does not specify any transport-layer protocols, while ISA-100.11a uses User Datagram Proto- col (UDP) and WirelessHART utilizes its own proprietary protocol. In application layer, HART-compliant protocol is used in WirelessHART to be compliant with existing HART devices. WIA-PA uses a proprietary protocol while ISA-100.11a does not specify any protocols in this layer.

2.4 MAC Protocols for WSN

MAC protocols of the data link layer in wireless communication are essentially used to determine the times during which a number of contending and colliding nodes access to a shared wireless medium. The problem solved by MAC protocols is a fundamental issue in any a wireless network. Thus, MAC protocols have been attracting a huge number of researchers to develop new MAC protocols for emerging wireless technologies during the past more than thirty years. When considering the choice of MAC protocols for a wireless technology, several factors have to be balanced: the feature of communication requirement, implementation complexity, antenna complexity, signal processing ability, and etc. This section will present the fundamentals of wireless MAC protocols, CSMA/CA protocol in detail, and MAC protocols in industrial automation standards.

2.4.1 Fundamentals of Wireless MAC Protocols

Traditionally, MAC protocols for wireless communications can be roughly classified [45] as: fixed assignment protocols, demand assignment protocols and random access protocols. Fixed assignment protocols: wireless bandwidth resources are divided and assigned to nodes or links in a long term. Each node or link uses its own dedicated resource without the risk of collisions. The typical protocols include Frequency Di- vision Multiple Access (FDMA), Code Division Multiple Access (CDMA), Space Di- vision Multiple Access (SDMA), and Time Division Multiple Access (TDMA). FDMA divides the available frequency band into a number of subchannels and assigns the subchannels to dedicated nodes which can transmit exclusively on their subchannels. It requires frequency synchronization, relatively narrow-band filters, and the ability of a receiver to tune to the channel used by a transmitter [45]. In CDMA systems, the narrow-band message signal is multiplied by a very large bandwidth signal called the spreading signal. The spreading signal is pseudo-noise code sequence that has a chip rate which is orders of magnitudes greater than the data rate of the message. The receiver in CDMA needs to know a codeword used by the transmitter. 18 Background and Related Work

Each user operates independently with no knowledge of the the other uses. SDMA controls the radiated energy for each user in space. SDMA serves different users by using spot beam antennas. These different areas covered by the antenna beam may be served by the same frequency or different frequencies. Sectorized antennas may be thought of as a primitive application of SDMA [46]. Due to the high implementation complexity, antenna complexity and/or sophisticated signal processing ability [45], FDMA, CDMA and SDMA has not been considered as the main technologies for WSN MAC protocols. However, TDMA has been utilized in industrial WSN since TDMA is able to provide a guaranteed access to the wireless channel and low complexity requirement to transceivers and microcontrollers. Demand assignment protocols: this type of protocols can be classified as centralized and distributed protocols. In centralized protocols, nodes send out requests for bandwidth allocation to a central node that either accepts or rejects the requests. In case of successful allocation, a confirmation is transmitted back to the request- ing node along with a description of the allocated resource. The node can use these resources exclusively. Resource deallocation is often done implicitly: when a node does not use its time slots any more, the central node can allocate it to others. An example of distributed demand assignment protocols are token-passing protocols. The right to initiate transmissions is tied to reception of a small special token frame. The token frame is rotated among nodes organized in a logical ring on top of a broadcast medium. Both of the two types of protocols require lots of energy consumption due to transceiver must be switched on most of the time. Therefore, they are not considered as appropriate protocols for WSN. Random access protocols: they are also known as contention based protocols. ALOHA and CSMA are typical examples. The basic idea of Slotted ALOHA is simple: let a wireless sender transmit at the beginning of the next transaction slot whenever they have packets to be sent. Slotted ALOHA exhibits poor performance because whenever one sender has a packet to transmit it does so without consid- eration of the others. CSMA improves Slotted ALOHA to transmit acyclic packets. The philosophy of CSMA is that when a sender generates a new packet, the channel is sensed, and if it is found to be idle, then the packet is transmitted. The channel sense scheme is generally known as clear channel assessment (CCA) in the IEEE wireless standards. The advantages of a contention-based protocol include: easy implementation; full bandwidth being available to any node; and suitability for use in dynamic networks without pre-scheduling. However, a contention-based protocol also has drawbacks: real-time communication cannot be guaranteed; fairness cannot be guaranteed because one node can occupy the medium for a long time; and the receiver must be active for the channel sense mechanism; therefore, it is difficult to determine the wake up time if the protocol allows sleep to save power. CSMA protocols are widely adopted in WSN protocols, while ALOHA protocols are also utilized in some special cases, e.g., shared slots in WirelessHART and ISA-100.11a standards. Considering the low energy consumption and resource constrained devices, the designs of MAC protocols for WSN generally focus on energy efficiency and low complexity. Abdelmalik et al. summarize WSN MAC protocols which regard energy efficiency as the prime design driver in [47]. The classifications include protocols 2.4 MAC Protocols for WSN 19

Figure 2.3: An example of the preamble sampling technique [47]

with common active period, preamble sampling protocols, scheduled protocols and hybrid protocols. Protocols with common active period: this type of protocols is an extension of conventional type of random access or contention based protocols. Nodes work in a duty-cycling mode with active/sleep periods. During active periods, nodes access to the wireless medium using contention based approaches. The typical protocols include SMAC (Sensor MAC) [48] and IEEE 802.15.4. The latter will be discussed in the next subsection. SMAC is briefly illustrated as follows. The primary design goal of SMAC is reducing energy consumption, meanwhile, it is also designed to support scalability and collision avoidance. Nodes are periodically enter active and sleep periods [49]. Nodes are locally time synchronized and neighboring nodes form virtual clusters to share a schedule, i.e., a common sleep and active periods, which is updated by a SYNC packet. Lacking of global synchronization, there are still collisions which are avoided by a carrier sense. In order to further reduce collisions (such as Hidden Terminal problem), SMAC utilizes request-to-send (RTS) and clear- to-send (CTS) prior to a series of short data packets transmissions and ACK. Besides, SMAC introduces a concept of message passing which fragment a long frame into short frames and transmit them in burst. The technique achieves energy saving by retransmitting corrupted short frames instead of a long frame, at the expense of un- fairness in medium access [49]. Preamble sampling protocols: in this approaches, a preamble sampling technique shown in Fig. 2.3 is utilized to further increase the proportion of sleep periods in a duty-cycling work mode. BMAC (Berkeley MAC) [50] and WiseMAC (Wireless Sensor MAC) [51] are typical protocols. BMAC is the origin of MAC implementation of TinyOS which is widely adopted in academic research. BMAC improve energy efficiency by accurately determine if the sensing channel is clear or not. As mentioned earlier, SMAC achieves collision avoidance by carrier sensing, which is realized by CCA. The common approach to implement CCA is to take single sample of received signal strength (RSS). If the value is larger than the noise floor, the channel is busy. Otherwise, the channel is clear. BMAC searches outliers in a certain samples. If a sample has significantly lower than noise floor during a sampling period, the channel is clear. WiseMAC improves energy efficiency by reducing preamble length in 20 Background and Related Work

the regular preamble sampling technique [52], where a wake up preamble of size equal to the sampling period is transmitted prior to every data frame so as to ensure the receiver is able to awake when data arrives [51]. WiseMAC enables every node to learn the sampling schedule information of other nodes so that a transmitter knows the wakeup time of the receiver. It is therefore able to shorten the preamble length and save energy. Scheduled protocols: this type of protocols is identical to the one of traditional fixed assignment protocols. Hybrid protocols combine the above protocols to achieve different objectives. ZMAC is a typical protocol in such category. ZMAC achieves high throughput and low latency under low contention using a contention based access approach, and achieved high throughput under high contention using a fixed assignment approach [53]. Thus, ZMAC achieves a high channel utilization with variable traffic patterns. As mentioned before, WSN in industrial automation, i.e., IWSAN, have real-time and reliable communication requirements on design of MAC protocols. Petcharat et al. explores to what extent existing MAC protocols for WSN can meet timely and reliable communication requirements [40]. The authors utilize SMAC as a reference point to evaluate delay and reliability of reviewed MAC protocols. The authors draw some conclusions as follows. Most existing WSN MAC protocols primarily focusing on energy consumption are unable to meet the requirement on both timely and reliable transmissions for mission critical applications. The authors addresses that WirelessHART is able to guarantee end-to-end delay and improve end-to-end reliability. Besides, Burst algorithm [54] and GinMAC protocol [55] have mechanisms to deliver the delay and reliability bounds for mission critical applications. The authors also reveal that only WirelessHART among reviewed protocols is evaluated in practical scenarios and the others are based theoretical studies which might be not realistic due to lacking of evaluations in real application scenarios. WirelessHART will be illustrated later and the burst algorithm and GinMAC are briefly discussed as follows. Burst algorithm introduces a new link-quality metric, Bmax , which is defined as the maximum number of time slots where packets transmission can fail due to a burst. After a long-term measurements (such as 21 days), links are classified into four categories: stationary links, asymptote-stationary links, epsilon-stationary links and non-stationary links, in which lower Bmax means higher stationary. Non-stationary links should be avoided for real-time transmissions; Epsilon-stationary links can be used for some real-time applications with some probability of missing deadline; Sta- tionary and asymptote-stationary links can be used for hard real time applications, addressed by the authors. The solution shows a better performance than widely used packet receipt rate. On top of that, the authors propose a least-burst-route to produce latency bounds of the streams considering link bursts and interference. Burst algorithm requires a pre-deployment measurement in plants and links quality is given based on the measurement, which is a limitation because of failing to capture short-term link dynamics due to movable obstacles or others. Suriyachai et al. have proposed GinMAC that considers timely and reliable data delivery by exclusively using guaranteed time slots and providing deterministic communications [55]. 2.4 MAC Protocols for WSN 21

The CSMA/CA protocols in general and MAC protocols in industrial automation standards will be illustrated in detail in the following subsections.

2.4.2 CSMA/CA Protocols

IEEE 802.11 is a widely accepted standard which adopts single channel mode. The fundamental access method of IEEE 802.11 MAC is a distributed coordination function (DCF) known as Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) protocol [56, 57]. Before transmitting a packet, a node operating in DCF scheme shall sense the channel to determine if another node is transmitting. If the channel is idle for a period of time equal to a distributed interframe space (DIFS), the node just starts transmitting. Otherwise, if the channel is busy, the node defers until the end of the current transmission. After the deferral, the node selects a random backoff interval and shall decrement the backoff interval counter while the medium is idle (this is the collision avoidance feature of the protocol), to reduce the probability of collision with packets being transmitted by other nodes. In order to further reduce collisions (such as Hidden Terminal problem), DCF utilizes a four-way handshaking mechanism, that is, exchanging short frames (known as RTS and CTS) prior to data transmission and ACK. In addition, to avoid channel capture, a node must wait a random backoff time between two consecutive new packet transmissions, even if the medium is sensed idle in the DIFS time. DCF adopts an exponential backoff scheme. At each packet transmission, the backoff time is uniformly chosen in the range ( 0, w − 1). The value w is called contention window, and depends on the number of transmissions failed for the packet. At the ﬁrst transmission attempt, w is set equal to a value CW min called minimum contention window. After each unsuccessful transmission, w is doubled, up to a m maximum value CW max = 2 CW min . IEEE 802.11 was originally designed to offer high throughput to wireless communications without taking energy consumption into account. Besides, it requires a high-performance MCU and a relatively high-complexity transceiver chip. There- fore, it is not considered suitable for applications of low power and low cost WSNs. Inspired by IEEE 802.11 MAC, IEEE 802.15.4 MAC also employs CSMA/CA as the fundamental access mechanism [2]. Compared with the former, CSMA/CA is simpliﬁed in IEEE 802.15.4, such as excluding the exchange mechanism of RTS and CTS control frames. Besides, some different schemes are adopted to support low- power mode, e.g., during the random backoff periods, nodes in the latter can enter sleep mode instead of sending the channel; the coordinator in the latter may enter sleep mode during the supported inactive portion of a superframe. IEEE 802.15.4 protocol has been analyzed and evaluated by [58–60] in detail. A brief description of the IEEE 802.15.4 MAC shall be given below. The IEEE 802.15.4 standard allows the optional use of a superframe structure. The superframe is bounded by network beacons sent by the coordinator and is divided into 16 equally sized slots. All transactions shall be completed by the time of the next network beacon. The beacons are used to synchronize the attached devices, 22 Background and Related Work

to identify the network, and to illustrate the superframe structure defined by a coordinator. The superframe can have an active and an inactive portion, as shown in Fig. 2.4. The active portion of each superframe is composed of three parts: a beacon, a contention access period (CAP) and a contention free period (CFP). The beacon shall be transmitted, without listen the channel or any backoff at the start of slot 0. The CAP period just starts after the beacon. Devices communicate during CAP period using a slotted CSMA/CA mechanism. The CFP, if present, follows immediately after the CAP and extends to the end of the active portion of the superframe. The network coordinator assigns so called guaranteed time slots (GTSs) in CFP to dedicated devices. Thus, there are no contentions for those devices to access the channel during CFP. Up to 7 GTSs can be assigned to devices in a network. The CSMA/CA algorithm in IEEE 802.15.4 is used before the transmission of data or MAC command frames transmitted within the CAP. The IEEE 802.15.4 uses two types of channel access mechanism: Unslotted CSMA/CA ( known as beacon based CSMA/CA) and slotted CSMA/CA (known as non-beacon based CSMA/CA). Unslotted CSMA/CA: When a device wishes to transmit a packet, it shall detect the channel using CCA after waiting for a random period. If the channel is idle, the device shall transmit its data. Otherwise, after the random backoff, the device shall wait for another random period before trying to access the channel again [2]. Slotted CSMA/CA: Beacon-enabled networks utilize slotted CSMA/CA mechanism. Unlike unslotted scheme, the backoff slots in the mechanism are aligned with the start of the beacon transmission. Each time a device wishes to transmit data frames during the CAP, it firstly locates the boundary of the next backoff slot, then waits for a random number of backoff slots. If the channel is busy, following this random backoff, the device shall wait for another random number of backoff slots before trying to access the channel again. Otherwise, the device can begin transmitting on the next available backoff slot boundary [2]. In both of the above cases, the CSMA/CA algorithm is implemented using units of time called backoff periods, where one backoff period shall be equal to a constant, i.e. aUnitBackoffPeriod (20 symbols) [2]. The CSMA/CA algorithm shall not be used for the transmission of beacon frames, acknowledgment frames, or data frames transmitted in the CFP. During the inactive period, devices can enter a low power mode for saving energy. One of the main considerations of the design of IEEE 802.15.4 CSMA/CA algorithm is to achieve energy efficiency. Energy consumption has attracted significant attentions in wireless sensor networks [61] and energy-efficient CSMA-based medium access control protocols have been well explored. But these protocols are unable to guarantee real-time transmissions, which is unacceptable in industrial applications with stringent timing requirements [33,62, 63]. As discussed above, IEEE 802.15.4 provides a GTS mechanism to allocate a specific duration within a superframe for time-critical transmissions. The authors in [64– 67] have proposed a series of adaptive scheduling or allocation mechanisms for GTS so as to achieve a low latency, improve bandwidth utilization or fairness for time critical messages. Zdenek et al. [68] have utilized a time division cluster schedul- 2.4 MAC Protocols for WSN 23

Figure 2.4: Illustration of IEEE 802.15.4 superframe structure [2]

ing mechanism to meet the deadlines while minimizing the energy consumption in beacon-enabled IEEE 802.15.4 network. Seong-eun et al. [69] have proposed a real-time ofﬂine message-scheduling algorithm which not only schedules periodic messages in GTSs but also conﬁgures parameters of a superframe to meet timing constraints. However, the standard and the above scheduling algorithms are based on a CSMA/CA dominant and hybrid mechanism with a limited number of GTS slots. Therefore, worldwide standards that meet the requirements in industrial automation need to be developed. MAC protocols of recent industrial standards shall be discussed as follows.

2.4.3 MAC Protocols in Industrial Standards

Industrial standards, WirelessHART, ISA-100.11a and WIA-PA, IEEE 802.15.4e leverage IEEE 802.15.4 physical layer with a low rate of maximum 250 kbps and a low implementation complexity for resource-constrained devices. They mainly utilize TDMA mechanism to provide guaranteed access to the wireless medium. An adaptive scheduling algorithm can effectively improve the QoS by means of delay optimization or mitigation [70, 71]. The standards also adopt packet-level channel hopping to improve the reliability of individual wireless links by combating external interference and multi-path fading. In 2.4 GHz IEEE 802.15.4 radio, there are 16 wireless frequency channels ranging from 11 to 26. WirelessHART (IEC 62591) MAC uses the TDMA access mechanism combined with channel hopping [5]. IEEE 802.15.4 superframe is not supported by the Wire- lessHART, where time is divided into consecutive time slots. A time slot in Wire- lessHART is a time unit that allows packet transaction once, which is similar to IEEE 802.15.4 GTS. Fig. 2.5 summarizes the WirelessHART packet format. A superframe in WirelessHART (called TDMA superframe and superframe refers to TDMA superframe unless otherwise statement in the rest of the thesis) is a collection of consecutive time slots shown in Fig. 2.6. Link transmissions are scheduled in the superframe and an adaptive scheduling algorithm can effectively improve QoS by means of delay optimization or mitigation [70, 71]. WirelessHART devices in a network share a channel hopping sequence determined by a centralized device called network manager. A 24 Background and Related Work

Figure 2.5: Summary of WirelessHART packet format

Figure 2.6: WirelessHART time slot and superframe [5]

channel blacklisting feature is provided to restrict the number of hopping sequence. Each link containing a transmitter and a receiver shall be assigned the channel hopping sequence and switch a channel after a transaction. ISA-100.11a (IEC 62734) MAC employs the same superframe structure. It supports three channel hopping operating modes: slotted channel hopping, slow chan- 2.4 MAC Protocols for WSN 25

nel hopping and hybrid combinations of slotted and slow hopping [6]. Just like WirelessHART, a centralized network manager is responsible to generate a hopping pattern. In slotted channel hopping mode, a time slot in a superframe is assigned a dedicated channel and the next time slot shall use the next successive channel in the hopping pattern. The hopping pattern is repeatable as time progresses. In slow hopping mode, multiple consecutive time slots are assigned a common channel. For a given hopping pattern, the standard provides a mapping between channel number in hopping pattern and MAC channel numbers. As applied to the 2.4 GHz DSSS IEEE Std 802.15.4 radio, channel numbers shall be limited to the range of 0 to 15, corresponding to IEEE Std 802.15.4 channel numbers 11 to 26. WIA-PA (IEC 62601) MAC has a compatible IEEE 802.15.4 superframe structure shown in Fig. 2.4, with redefined functions for CAP, CFP and inactive period. The CAP is used for packet transmission of device joining, intra-cluster management and retransmissions. The access method in the CAP is CSMA/CA mechanism defined in IEEE 802.15.4, which is different from WirelessHART and ISA-100.11a. Each WIA- PA time slot duration in CFP is configurable and compatible with IEEE 802.15.4 GTS. Except the sleep mode, intra- and inter-cluster communication is allowed during the inactive period. WIA-PA supports three channel hopping mechanisms: adaptive frequency switch (AFS), adaptive frequency hopping (AFH), time slot hopping (TH) in which channel is changed per time slot. The channel hopping sequence is determined by the network manager. IEEE 802.15.4e MAC defines a MAC amendment to the existing standard IEEE 802.15.4-2006 to better support industrial markets, as mentioned before. All motes in an IEEE802.15.4e network are synchronized and time is split into time slots, each typically 10 ms long. Time slots are grouped into a super-frame which continuously repeats over time. Each device in the network obtains its schedule from a distributed scheduling algorithm. The schedule then instructs the device what to do in each time slot: send, receive, or sleep. Channel diversity is obtained by specifying, for each send and receive slot, a channel offset. The same channel offset translates into a different frequency on which to communicate at each iteration of the super-frame. The resulting channel-hopping communication reduces the impact of external interference and multipath fading, thereby increasing the reliability of the network [13].

2.4.4 Open Issues

As mentioned in Chapter 1, there are different traffic requirements with different critical classes. For example, the three categories of data traffic, safety, control and monitoring, have critical classes in descending order. ISA 100.11a, WirelessHART and WIA-PA provide a limited service differentiation in MAC schemes. They support four traffic categories which are defined from the highest priority to the lowest priority as: command, process-data, normal and alarm. High-priority packets are allowed to access the medium prior to low-priority packets. The deterministic and synchronous multi-channel extension (DSME) mechanism in IEEE 802.15.4e is one of the important extensions to IEEE 802.15.4. DSME supports prioritized channel access by reserving specific DSME guaranteed time slots for high priority traffic. 26 Background and Related Work

However, both of the priority mechanisms do not mean that a high-priority packet at a device can immediately be sent out after its generation. Each device or link has its own dedicated slot and an immediate transmission may occupy another device’s slot thus causing a high risk of collisions. Thus, a high-priority packet has to wait for the next assigned or reserved slot after the generation of the packet, which can be after a number of slots and which may be even longer in case of retransmissions. Exceeding the required delay bound for unpredictable and emergency traffic could lead to system instability, economic and material losses, system failure and, ultimately, a threat to human safety. Guaranteeing the timely delivery of the IWSAN critical traffic and its prioritization over regular traffic (e.g. non-critical monitoring traffic) is a significant challenge. A priority based MAC protocol for IWSAN, therefore, need to be developed. Besides, another challenge for the IWSAN is that: non-compatible channel hopping patterns, routing protocols and upper-layer protocols in WirelessHART, ISA- 100.11a, WIA-PA, IEEE 802.15.4e (supported by different global suppliers in the market) may incur a poor performance regarding to coexistence and interoperability among devices using different standards, although these standards use the same radio in almost an identical way to get long battery life, guaranteed medium access for real-time communication and avoidance of multipath fading via channel hopping.

2.5 TDMA Scheduling

As mentioned, scheduled MAC protocols assigns divided wireless bandwidth to nodes or links in a long term, and each node or link uses its own dedicated resource without the risk of collisions. Thus, scheduled protocols are generally suitable for periodic and high-load trafﬁc. Industrial standards in industrial automation introduces scheduled protocols, such as TDMA, to provide guaranteed access or bounded latency to the wireless medium in order to meet the real-time transmission requirement. In addition, TDMA is attractive because - once the schedule is set up - there are no collisions, no overhearing, and minimized idle listening [47]. Therefore, scheduling algorithms play an important role for TDMA-based wireless networks.

2.5.1 Scheduling Problems

Scheduling algorithms are inherently designed for resource allocation problems. Examples are scheduling algorithms for meeting timing constraints in real-time operating systems and scheduling mechanisms for railway transportation. Scheduling problems are not new in wireless networks. The early problems and solutions can be traced back to the mobile radio system: users are scheduled by a base station on different traffic channels depending on their traffic requirements and channel conditions [72]. Scheduling algorithms in wireless multi-hop is more challenging. Existing TDMA scheduling approaches [73–77] define the goal of TDMA scheduling in finding the minimum number of slots required to schedule link transmissions. 2.5 TDMA Scheduling 27

In these papers, the authors reduce the number of slots by means of spatial reuse. As a consequence, conflict-free links can be assigned to the same slot. Neverthe- less, the presented algorithms do not account for inbound-outbound delay incurred by TDMA scheduling, given that certain inbound links are scheduled after their re- spective outbound links. Thus, minimum length schedules do not automatically guarantee a minimal end-to-end delay. In [78], Djukic and Valaee introduce a general TDMA scheduling problem for multi-hop wireless networks. The authors derive a worst-case inbound-outbound delay in terms of link transmission orders and discuss min-max scheduling delay optimization based on a conflict-free TDMA schedules, using an improved Bellman- Ford shortest path algorithm to find optimal round-trip paths. The algorithms show good performance properties for delay and bandwidth efficiency. The algorithms in the paper mainly targets TDMA scheduling over IEEE 802.16 and IEEE 802.11s based wireless networks. Huan Li et al. in [79] propose a message scheduling algorithm, which provides timeliness guarantees for multi-hop real-time transmissions for a team of robots equipped with sensors. The main wireless technology is based on IEEE 802.11. A proprietary MAC method is proposed together with the scheduling algorithm, instead of CSMA/CA based MAC. The TDMA-based scheduling problems have been attracting researchers in both IEEE 802.16, IEEE 802.11s or IEEE 802.11 based wireless network, and researchers in IEEE 802.15.4 based WSN. Al- though both of IEEE 802.x and IEEE 802.15.4 based scheduling problems have similar network models and conflict models, the scheduling problems and solutions are different because of different data transmission requirement and network topologies. The former normally has high data rates and large packets, thus, continuously sub- sequent slots need to be assigned to a single pair of transmitter and receiver. There are periodic data transmissions with low data rates and small packets in the latter case so that each transaction only occupies a single slot. In [80], Ergen and Varaiya present TDMA scheduling problem for WSN to determine the shortest length conflict-free assignment of slots in a superframe, during which the generated packets from each node reach their destination. The authors prove this problem to be NP-complete and the authors propose a level-based scheduling algorithm to solve it. The remainder of the section will discuss network model, wireless interference model and scheduling algorithms in detail.

2.5.2 Network Model and Wireless Conﬂict Model

In order to solve the scheduling problems, the TDMA based network is normally modeled as a directed connectivity graph G(V, E ), where the vertices V = {v0, · · · , v n} are the set of nodes with v0 denoting the Gateway or Sink. The edges between vertices E = {e1, · · · , e m} represent wireless communication links. The research community has only recently developed practical full-duplex radios [81]. Therefore, the half-duplex radio is considered throughput the thesis in which a transmitter cannot transmit and receive at the same time. 28 Background and Related Work

Table 2.4: Notations of network model for scheduling algorithms

V vertices that represent a set of nodes in the network

E edges that represent wireless links

C edges that represent conﬂict links

G(V, E ) directed connectivity graph that model the network

G(V, C ) a graph that models the network with nodes and conﬂict links

G(V, E, C ) a graph that model the network with nodes, links, and conﬂict links

Spatial reuse is one of the mechanisms used by scheduling algorithms to mitigate end-to-end delay in TDMA-based multi-hop networks. As a consequence of spatial reuse, conflict-free links can be assigned to the same slot, while links whose transmissions may conflict with each other cannot be allocated to the same time slot. Two sources for wireless transmission conflicts can be distinguished, namely primary conflict and secondary conflict [78, 80]. Primary conflict occurs when one node performs more than one action in a single time slot, i.e., receiving and transmitting messages, or concurrently receiving from different transmitters. Secondary conflict can occur if a receiver is within the radio range of multiple transmitters, but is solely interested in one of them.

Therefore, a conflict graph is necessary to model. Let edges C = {c1, · · · , c k} ⊂ E represent conflicting wireless links. Then, G(V, C ) can be used to model the network with nodes and conflict links, and G(V, E, C ) represents the network with nodes, links, and conflict links. The notations of the network model is shown in Table 2.4.

2.5.3 Scheduling Algorithms

Existing scheduling algorithms in WSN can be roughly classified into scheduling for broadcast traffic and scheduling for unicast traffic according to different traffic types. Data collection is a fundamental traffic type in WSN where each sensor node periodically transmits packets containing measured or sensed data to the destination node, i.e., Sink or Gateway. In [47], the authors classify scheduling methods in the context of WSNs as: scheduling of links (which consist of wireless transmitter and receiver pairs), scheduling of transmitter and scheduling of receivers . The category is made according to which a scheduling approach assign a dedicated slot: each link, transmitter or receiver in a network. Since transmitter and receiver know exactly when to wake up in the first approach, this is the most energy efficient solution, which have been attracting lots of researchers. In order to illustrate how a scheduling algorithm works, a classic scheduling al- 2.5 TDMA Scheduling 29

Table 2.5: Notations of network model for the level-based algorithm

L a set of tree levels; the tree level of a node is the depth of the node in the tree

VL vertices that represent a set of virtual “nodes”; a set of nodes in a same tree level is regarded as a single virtual “node”

EL edges that represent a set of virtual “links” between virtual “nodes”

CL edges that represent conﬂict virtual “links”

G(VL, E L, C L) a virtual “graph” that model the network with virtual “nodes”, virtual “link”, and virtual conﬂict “links”

Algorithm 1 Construct a virtual “graph” consisting of virtual “nodes”, “links” and conﬂict “links” Input: V , E, and n which is the number of levels in the tree. Output: VL, EL, and CL 1: construct VL = {vL1, · · · , v Ln }, in which each element represent a virtual “node”, i.e., a set of nodes in the corresponding tree level. 2: construct EL, in which each element is a virtual “link” between two adjacent virtual “nodes” in VL. 3: for i ← 1 to n do 4: if ∃ (u, v ) ∈ CL, satisfying node u is at level i and node v is at a level j which is smaller than i. then 5: add edge (vLi , v Lj ) to CL

gorithm as an example will be discussed in detail as follows. The algorithm, which adopts the method of scheduling of unicast links, is called level-based algorithm proposed by Sinem et al. in [80] for data collection in WSN. It is the main comparison reference for evaluation in the Paper III of the contributed publications. The algorithm descriptions [80] have been re-organized here. The concepts of virtual “nodes”, virtual “links” and a virtual “graph” have been introduced to differentiate them from nodes in the networks, links between nodes and a graph consisting of nodes and links. A virtual “node” is a set of nodes with the same tree level (depth). The notations for the algorithm is shown in Table 2.5. The algorithm assumes that a tree topology is used for data collection. It has three parts. Firstly, Algorithm 1 classiﬁes nodes into different level sets L according to their depths in the tree, and construct a virtual “graph” by generating virtual “nodes” (VL), virtual “links” (EL) and conﬂict virtual “links” (CL). Secondly, Algorithm 2 assigns colors to the virtual “nodes” (VL). Lastly, Algorithm 3 determines the slots for each color group until all nodes are able 30 Background and Related Work

Algorithm 2 Color assignment for the virtual “nodes” in the virtual “graph” Input: VL, EL, and CL Output: {(vL1, c 1), · · · , (vLn , c n)}, where, ci ∈ { 1, 2, 3, ..., M } and M is the number of colors. 1: order the “nodes” as {k1, k 2, ..., k n}. 2: for l ← 1 to n do 3: i ← 1 4: while ∃ j assigned to color i, satisfying (j, k l) ∈ CL do 5: i ← i + 1 6: assign color i to kl

Algorithm 3 Color assignment for nodes in the graph G(V, E ) Input: G(V, E ), G(V, C ), and color assignment for the virtual “nodes” with M colors Output: schedule for nodes of G(V, E ) 1: while at least one packet has not reached Gateway do 2: for l ← 1 to M do 3: construct a temporary set SET s containing “nodes” which have a color s 4: construct T , T ← ∅ 5: for j ← 1 to kSET sk do 6: find a nonconflicting set of nodes ∈ SET s(j) {a virtual “node” represents a set of nodes. } 7: T ← T ∪ the nonconflicting set 8: if T 6= ∅ then 9: construct a temporary set SET os containing “nodes” not corresponding to color s 10: for each node k belonging to a virtual “node” ∈ SET os do 11: if (k, j ) 6∈ C, ∀ j ∈ T then 12: T ← T ∪ k 13: assign current slot to set T to send one packet to the sink.

2.5.4 Open Issues

The majority of existing TDMA scheduling approaches [58, 73–77, 80, 82–84] do not consider the effects of unreliable wireless links on scheduling algorithms. Link failures can result in schedules of broken order. Individual wireless links are inherently unreliable. Observations on real-platform testbeds with low-power wireless sensor networks show that links have wide ranges of packet delivery ratio which 2.6 Routing Protocols in WSN 31

can vary significantly over time, even without intra-collisions or high portions of external interference [11–13]. The dynamics of individual links are even more unpredictable in industrial environments, because of the human blockage, machinery and products obstacles. Furthermore, the majority of existing TDMA scheduling algorithms do not consider the configurational overhead. A centralized scheduling mechanism for industrial applications is thus taken as an example to describe this issue. In centralized scheduling, schedule requests come with topology and routing information when sent to the sink. The sink then runs a centralized scheduling algorithm and distributes the schedules to the sensors. Neighbors and routing information may change very frequently because of the unreliable nature of the individual wireless links and obstacles, thus generating heavy configurational overhead. This overhead in return increases the end-to-end delay.

2.6 Routing Protocols in WSN

The routing function in network layer is important for a multi-hop network. The routing algorithms are essential used to determine good paths consisting of a series of routers from a source node to an end-to-end destination node. This section is not intended to comprehensively discuss about routing protocols in WSN. Instead, previous milestone protocols in WSN will be studied in brief. Due to the complexity and importance, routing algorithms have been attracting thousands of researchers in WSN. The routing algorithms can be classified into link-state algorithms and distance-vector algorithms according to whether they are centralized or decentralized [85]. Previous work on routing protocols in WSN addressed various objectives according to the specific applications. Flooding and gossiping protocols are early solutions for relaying data in sensor networks [86]. They do not need any routing algorithms and topology maintenance. This type of solutions is particularly useful for relaying data for small groups of mobile nodes. Geographic routing protocols are location- awareness solutions [87], which may combine energy awareness [88] for sensor networks. Cluster-based hierarchical routing is also an important type for routing solutions in WSN. It influences the design of the cluster tree network architecture in IEEE 802.15.4. Fig. 2.7 shows an example given by IEEE 802.15.4. Nodes are grouped into several clusters. In each cluster, there is one coordinator providing synchronization services. Among these coordinators, only one of them can be selected as the overall coordinator, which may have greater computational resource than any other nodes. A typical and well-known approach is Low-Energy Adaptive Clustering Hierarchy (LEACH) [89]. LEACH is proposed for building an energy-efficient and low-latency architecture combining routing and MAC in remote monitoring applications. A distributed self-organization technique and routing protocols are proposed for forming the clusters, and intra-clusters and inter-clusters communications. TDMA is used for intra-cluster communications, while CDMA is utilized for inter-cluster communications. Other cluster-based routing protocols include the Threshold sensitive Energy 32 Background and Related Work

Figure 2.7: An example of cluster tree network architecture in IEEE 802.15.4 [5]

Efficient sensor Network protocol (TEEN) [90], the Hybrid, Energy-efficient, Dis- tributed clustering protocol (HEED) [91] and so on. Directed diffusion is regarded as a milestone in data-centric routing research in WSN [86]. In order to save energy, dif- fusing data schemes are used to reduce complexity of routing. Based on the idea of directed diffusion, a Gradient Based Routing (GBR) [92] has been proposed. Gradi- ents in GBR are established by flooding some a type of information for a sink, called “interest” message, in the network and each node finds a minimum hops to the sink. There are a series of routing protocols which further develop gradient-based routing such as GRAdient Broadcast (GRAB) [93]. Please see the routing survey articles in [86,94, 95] for more detailed reviews on routing algorithms in WSN. Mobile Ad-hoc Networks (MANET) working group in IETF has contributed several well-known routing protocols for wireless multihop ad-hoc networks. These protocols include: Ad hoc On-Demand Distance Vector (AODV) routing protocol (2003) [96], Optimized Link State Routing protocol (OLSR) (2003) [97], Topology Dis- semination Based on Reverse-Path Forwarding (TBRPF) (2004) [98], The Dynamic Source Routing protocol (DSR) for Mobile Ad Hoc Networks for IPv4 (2007) [99]. They can be classified as reactive/on-demand protocols and proactive/table-driven protocols. AODV and DSR are reactive distance-vector protocols, while OLSR and TBRPF are proactive link-state protocols. The MANETs have mainly been described as [100]: an autonomous system consisting of mobile nodes; multi-hop dynamic topologies which may randomly and rapidly change at unpredictable times; bandwidth constrained, variable capacity link; energy constrained operation; Limited physical security. These characteristics fit some early WSN applications, thus, many 2.6 Routing Protocols in WSN 33

protocols in WSN “embrace” MANET routing protocols, while, in turn, WSN widely extends the applications of MANET. For example, Zigbee adopts AODV as its main routing protocols. Nevertheless, the needs of considering realistic applications for the research work of WSN in recent years, the initial vision changed. In most realistic applications such as industrial automation applications, most of wireless sensors or actuators are static or with low mobility. Most of routing protocols in WSN are evaluated in simulators or limited experimental tests. There are a series of practical protocols which have been widely experimented in multitude of test-beds. MintRoute [101], Collection Tree Protocol (CTP) [102] and TinyRPL [103] have been adopted as de-facto routing protocols in different versions of TinyOS which is the first open-source operating system designed for WSN. CTP has been implemented in different test-beds and simulators and have been widely used in research, teaching, and in commercial products. Pa- per III and IV among contributed papers of this research work have used CTP as the routing protocol. Thus, a brief introduction will be given in this section. The routing protocols in WSN have been well explored. Recent years, IPv6-based WSN routing protocols have been attracting a lot of researchers [104] in this area. The IETF 6LoWPAN working group has started in 2007 to work on specifications for transmitting IPv6 over IEEE 802.15.4 networks. On top of that, the IETF chartered ROLL working group to standardize a link-independent IPv6 routing protocol for resource constrained devices [105]. The RPL (pronounced Ripple) [95] has been defined to build a robust topology over lossy links with minimal state requirements. RPL has been implemented in widely used open-source platforms such as TinyOS and Contiki [106]. Besides, as the size of network scale in IWSAN increases, emerging routing protocols which satisfy real-time requirement in IWSAN become more and more significant. In [107], Yanjun Li et al. proposes a two-hop information based routing which maps packet deadlines to a velocity and can reduce packet deadline miss ratio. In [108], the authors propose a gradient routing with two-hop information, which combines a two-hop velocity-based routing algorithm in [107] and gradient routing to reduce delays and energy consumption. In [109], Kan Yu et al. propose a controlled flooding based real-time routing scheme for some industrial control applications.

2.6.1 Collection Tree Protocol

Collection Tree Protocol (CTP) [102], a well-known protocol in WSN, is distance vector based protocol designed for WSN. The distance vector algorithm is iterative, asynchronous, and distributed routing algorithm [85]. It is distributed in that each node without global information, receives information from one or more of its neighbors, performs a calculation, and then distributes the results to its neighbors. It is iterative in that this process continues on until no more information is exchanged between neighbors. The algorithm is asynchronous in that it does not require all of the nodes to operate in lockstep with 34 Background and Related Work

Figure 2.8: Illustration of the operation of CTP protocol

each other.

Let dx(y) be the cost of the least-cost path from node x to y. Then the least costs are related by the celebrated Bellman-Ford equation, namely,

dx(y) = min v{c(x, v ) + dv(y)}, (2.1)

where the min v is taken over all of x’s neighbors. After traveling from x to v, if we then take the least-cost path from v to y, the path cost will be c(x, v ) + dv(y). Since we must begin by traveling to some neighbor v, the least cost from x to y is the minimum of c(x, v ) + dv(y) taken over all neighbors v. CTP computes the routes from each node in the network to Sink in the network. CTP uses the following mechanisms to overcome the challenges in a highly dynamic wireless network. It uses adaptive beacon broadcast messages to form and maintain a spanning tree topology. An adaptive beacon mechanism realized by a so called Trickle algorithm [110] is adopted. When topological inconsistencies arise, beacons are sent more frequently [102]. Otherwise, the beaconing rate is decreased exponentially. Speciﬁcally, in the absence of topological changes, the beaconing interval is regularly doubled until it reaches a maximum value which triggers only a few beacons per hour. Upon topological changes, the interval is reduced to allow for fast gradient re-convergence. Fig. 2.8 illustrates the operation of CTP protocol. Every node maintains an estimate of the cost of its route to a collection point. The expected transmission count (ETX) is the default cost metric, but any similar gradient metric can work just as well. 2.7 Implementation of Communication Protocols in WSN 35

Figure 2.9: A typical commercial development stack for WSNs

A node’s cost is the cost of its next hop plus the cost of its link to the next hop: the cost of a route is the sum of the costs of its links. Collection points (Sinks) advertise a cost of zero. Each data packet contains the transmitter’s local cost estimate. When a node receives a packet to forward, it compares the transmitter’s cost to its own. Since cost must always decrease, if a transmitter’s advertised cost is not greater than the receiver’s, then the transmitter’s topology information is stale and there may be a routing loop. Using the data path to validate the topology in this way allows a protocol to detect possible loops on the ﬁrst data packet after they occur.

2.7 Implementation of Communication Protocols in WSN

The methods for implementation of communication protocols in realistic testbed can be roughly classiﬁed into commercial development kits and open source. Fig. 2.9 presents a typical commercial development kit for software protocol stack. A real- time operating system, PowerPac, based architecture with multi-processors support and ARM Cortex-M3 based hardware constitute the test-bed platform. IAR Em- bedded Workbench [111] (one of the most commonly used integrated development environments for embedded systems) and JTAG debugging tool have been used for 36 Background and Related Work

programming and debugging. The ST Simple MAC in the figure is a commercial MAC protocol that realizes basic IEEE 802.15.4 CSMA/CA (which can be disabled) mechanism. As can be seen, the black box of the operating system and stack protocol, on the one hand, reduces developers’ workload, and, on the other hand, degrades execution efficiency and flexibility. Two well-known open-source based operating systems in WSNs will be discussed in the following. TinyOS is an event-driven open-source operating system designed for low power and low cost wireless devices [14] (e.g., microcontroller: MSP430 with 48 kB of ROM and 10 kB of RAM; low-power radio chip: CC2420 compatible with IEEE 802.15.4). It has been widely adopted by thousands of researchers worldwide in wireless sensor network community. On the one hand, TinyOS provides a series of principles, techniques, software structures and mechanisms to achieve computationally efficient programming/coding on resource limited embedded devices. For example, the principle of minimizing resource use is able to achieve efficiently and flexibly minimizing the RAM and code ROM. On the other hand, it provides code modules to run on a range of hardware platforms (e.g., Telosb with MSP430 and CC2420). The code modules include schedulers, timers, power managements, hardware drivers, radio stacks, network protocols and so on. The CC2420 radio stack in TinyOS is non-synchronized CSMA/CA based MAC protocol in practice. A lot of researchers have been attracting to use TinyOS to develop protocols and contribute their codes. There is 25,000 downloads per year in average during the last decade [112]. For example, the Flooding Time Synchronization Protocol (FTSP) [113], a well-known multi-hop time synchronization protocol for WSNs, is a typical contribution from researchers outside the core developers of TinyOS community. Another example is Berkeley OpenWSN [17] which have primarily implemented their proposed communication protocols in TinyOS. Berkeley OpenWSN is an open- source stack intending to implement low-power wireless standards such as IEEE 802.15.4e, 6LoWPAN, RPL and CoAP. The implementation of time synchronization module for this research work has absorbed the methods in FTSP and OpenWSN. TOSSIM is a default simulator in TinyOS. It emulates the behavior of underlying raw hardware by replacing several low-level hardware components with software equivalents [18]. Thus, the transition from the simulation environment to a real platform is relatively easy. The use of TOSSIM has another advantage. Its radio model has the ability to capture the short-term link dynamics of low-power radio, by accurately simulating real-world 2.4 GHz environmental noise. It simulates this noise using the closest-fit pattern matching model [60]. TOSSIM inherently supports one single wireless channel. Contiki is an another well-known open-source for WSN and recently it is an- nounced to support a wide range of smart objects [23] for Internet of Things. It is built around an event-driven kernel and provides optional preemptive multi- threading that can be applied to individual processes [106]. The most outstand- ing contributions are designs of IPv6 stacks such as µIP [114], ContikiRPL [104] for resource-constrained devices. Chapter 3

Summary of Contributed Papers

This chapter summarizes the contributed papers by addressing the achievements and limitations. Table 3.1 shows the objective, topics and my participation associated with the corresponding contributed papers. The detailed summary will be discussed as follows.

Paper I PriorityMAC: A Priority-Enhanced MAC Protocol for Crit- ical Trafﬁc in Industrial Wireless Sensor and Actuator Networks

In this article, a vital issue of IWSAN, namely, service differentiation for priority of critical traffic over non critical traffic, is addressed in real-time industrial automation applications. Exceeding the required delay bound for unpredictable and emergency traffic could lead to system instability, economic and material losses, system failure and, ultimately, a threat to human safety. However, guaranteeing the timely delivery of the IWSAN critical traffic and its prioritization over regular traffic (e.g., noncritical monitoring traffic) is a significant challenge. The paper proposes Priority- MAC, a Priority enhanced MAC protocol, designed for critical traffic in IWSANs, and the first priority enhanced MAC protocol compatible with IWSAN industrial standards. We present the design, implementation, performance analysis and evaluation of PriorityMAC. This paper has four original contributions. Firstly, our protocol provides a service differentiation in wireless medium access for traffic categories ( T C 1, T C 2, T C 3 and T C 4) of different priorities. It utilizes the industrial standards, WirelessHART and

37 38 Summary of Contributed Papers

Table 3.1: Objective, topics and my participation associated with corresponding papers

Objective Topics Papers My participation

How to enable I am fully responsible for ideas, unpredictable implementation and experiments in critical traffic to testbed; I am also responsible for most of problem description, perfor- timely delivery Paper I and to have mance analysis, simulation evaluation, experimental evaluation, writ- Service priority over ing, and revisions according to com- differentiation real-time ments from reviewers. for guaranteed non- TDMA-based or less-critical wireless traffic? I merely contributed to test-bed im- medium How to improve Paper II plementation and corresponding per- access the transmission formance evaluation. efficiency for burst acyclic I am fully responsible for ideas, traffic in implementation and experiments in WirelessHART Paper V testbed; I am also responsible for most and ISA100.11a of problem description, experimental based IWSAN? evaluation and writing.

I am fully responsible for ideas, implementation and simulations in Paper III TOSSIM; I am also responsible for most of the theoretical analysis, evalu- How to adapt to ation, writing and revisions according Adapt to link link dynamics to comments from reviewers. dynamics for for scheduling scheduling algorithms in algorithms realistic I am fully responsible for ideas, im- environment? plementation and experiments; I am Paper IV also responsible for most of problem description, performance evaluation and writing.

ISA100.11a, as a baseline and employs a novel medium access control scheme to enhance access for the critical and aperiodic trafﬁc. Secondly, we provide a theoretical analysis of average access delay and utilization of wireless resource for different trafﬁc priorities, which closely matches with the corresponding simulations. PriorityMAC achieves a high channel utilization Summary of Contributed Papers 39

3 10

2 10

Delay(ms) 1 10

deferred TC4 TC3 TC2 TC1 0 10 0 2 4 6 8 10 12 14 16 Time(hour)

Figure 3.1: PriorityMAC experiment shows it is able to efﬁciently handle different categories with different latency requirements.

and low latency under low transmission rates of periodic traffic, because of the high probability in relation to the hijacking of idle slots for periodic traffic. It achieves low latency for delay-sensitive and unpredictable traffic at the expense of a slight decrease in channel utilization under high transmission rates of periodic traffic. This is because the high probability of deference or even damage of periodic traffic may result in a decrease of channel utilization. The effects on channel utilization of Priority- MAC become negligible when the transmission probabilities of T C 1 and T C 2 are small, while PriorityMAC achieves a significant improvement in latency. Thirdly, PriorityMAC is implemented using TinyOS and is evaluated on a testbed of 40 TelosB motes. Telosb mote uses a microcontroller of MSP430 with 48kB of ROM and 10kB of RAM and a IEEE 802.15.4-compliant low-power radio chip, CC2420. Lastly, PriorityMAC achieves a significant reduction regarding the latency in experimental evaluation. Fig. 3.1 shows a delay experiment of T C 1, T C 2, T C 3, and deferred T C 4 (due to an arrival of a higher priority packet prior to T C 4) within 16 hours. The T C 4 devices which transmit T C 4 packets changed the data refresh interval every 2 hours during the experiment. The figure indicates that the proposed solution can efficiently handle different categories with different latency requirements. The experimental comparison with the priority access methods in the current industrial standards shows that PriorityMAC achieves a reduction in latency of 94% for the highest priority traffic and 93% for secondary priority traffic, on average. The reduction is achieved at the expense of only a 0.17% increase in third priority traffic access delay and a 0.19% deferring ratio from the lowest-priority packets. Limitations: PriorityMAC supports four types of access methods concurrently operating within the protocol. Each access method corresponds to one traffic category. From a theoretical standpoint, the traffic categories appear to be arbitrary. Furthermore, there is a limitation about the highest priority traffic, that is, no collision of two or more of this type of packet is assumed. 40 Summary of Contributed Papers

Paper II CCA-Embedded TDMA enabling Acyclic Trafﬁc in In- dustrial Wireless Sensor Networks

In this article, we propose a CCA-Embedded TDMA scheme that improves the transmission efficiency and system stability of a TDMA-based system that generates burst acyclic traffic, such as the WirelessHART protocol and ISA-100.11a. A Markov model is provided to analyze the performance of CCA-Embedded TDMA. Further- more, we have implemented CCA-Embedded TDMA in a testbed composed of 10 Telosb devices using TinyOS. I merely contributed to the test-bed implementation and the corresponding performance evaluation. Thus, the summary here only cov- ers this contribution. To evaluate the performance of the proposed CCA Embedded TDMA scheme, we deployed 10 field devices and a gateway in an indoor environment. The field devices were deployed in a circle around the gateway with a distance of 2.5 m between the gateway and field devices. All of the transmissions are on the same channel. In the start-up phase, all of the field devices are notified of the frame generation rate upon receiving the beacon frames. Then, each field device generates and transmits 94-byte frames for 5 min during a shared transaction slot. For every device, if the current frame has not been sent out, no new frame is generated. This design guarantees a single buffer system. The obtained results show that CCA-Embedded TDMA improves the throughput compared with the WirelessHART MAC. Limitation: The limited number of devices used in the experiment and the lack of multihop network scenarios are the main limitations of this article.

Paper III SAS-TDMA: A Source Aware Scheduling Algorithm for Real-Time Communication in Industrial Wireless Sensor Networks

Scheduling algorithms play an important role for TDMA-based wireless sensor networks. Existing TDMA scheduling algorithms address a multitude of objectives. However, their adaptation to the dynamics of a realistic wireless sensor network has not been investigated in a satisfactory manner. This is a key issue considering the challenges within industrial applications for wireless sensor networks, given the time-constraints and harsh environments. In response to those challenges, we present SAS-TDMA, a source-aware scheduling algorithm in this article. It is a cross-layer solution which adapts itself to network dynamics. It realizes a trade-off between scheduling length and its conﬁgurational overhead incurred by rapid responses to routes changes. We implemented a TDMA stack instead of the default CSMA stack and introduced a cross-layer for scheduling in TOSSIM, the TinyOS simulator. TOSSIM makes the program code transition

Summary of Contributed Papers 41

' ( ( ) * + , - . / 0 1 2

) 9 : * + , - ; 0 1 < / 0 1 2

3 4 5 6 7 8 6 5 6

* ( ( * + , - = ; 0 1 2

Figure 3.2: A comparison of nodes proportions with different PDR levels for SAS-TDMA algorithm and level-based algorithm with the refresh interval of 3.1s.

between the simulation and real platform by only replacing a small number of low- level components. It also has the ability to capture the link dynamics by simulating the real-world noises. The evaluation in our work is thus close to the experiments in a real system. Fig. 3.2 shows the performance comparison of the proposed solution with a classic scheduling algorithm, the level-based algorithm. During the simulation, we de- liberately uses a heavy noise so that the PDR of each individual link was poor. The retransmission attempts stopped when hitting an upper bound of delay. Nodes were said to have poor PDR level if PDR was below 30%, intermediate between 30 and 90%, and good otherwise. The ﬁgure indicates that the performance of the level- based algorithm decreases by about 30% for good PDR compared with our solution. Numerical results show that the SAS-TDMA improves the quality of service for the entire network. It achieves signiﬁcant improvements for realistic dynamic wireless sensor networks when compared to existing scheduling algorithms with the aim of minimizing the latency for real-time communication. Limitations: There are still differences between TOSSIM simulation and the real platforms. The implementation of SAS-TDMA in TOSSIM did not include the time synchronization. Additionally, the proposed SAS-TDMA was evaluated in the network based on the CTP for constructing and maintaining a routing tree. The effects of other routing protocols, such as RPL, on the performance of SAS-TDMA need to be further investigated. 42 Summary of Contributed Papers

Paper IV Distributed Data Gathering Scheduling Protocol for Wire- less Sensor Actor and Actuator Networks

In this paper, we present a novel cross-layer scheduling protocol for data gathering transmission. In WSAN, there is typically a short-length sensor data with a relatively large overhead in each packet. Data gathering transmission is able to improve the QoS of the whole network by means of efficiently scheduling packets transmission and combining packets from multiple sources and then forwarding a new packet. There are two key issues which have not been dealt with in a satisfactory manner for the previous work involving data gathering transmission for sensor networks. Firstly, most of the solutions have not considered the case involving topology changes and retransmission in realistic wireless networks. Wireless sensor and actuator networks undergo frequent topology changes due to physical environmental changes, node join, node failures, and time-varying channel conditions. An individual wireless link is inherently unreliable. It has been observed from using real- platform testbeds with a low power wireless network that links have a wide range of packet reception ratios, which can vary significantly over time even without intra- collisions and heavy external interference. It is even worse for industrial applications in harsh environments as both humans and machinery alter the RF environment. In addition, the case involving retransmission due to unreliable and asymmetric wireless links has to be considered when designing the protocols. Secondly, most of the previous gathering scheduling algorithms have not considered another important problem: the scheduling configuration overhead. The previous work needs to maintain one-hop and two-hop neighbors. Each node has to construct its competitor set from two neighbors and has to make decisions regarding its schedule by itself.We call these algorithms as node dominant decision scheduling. They only consider the delay after scheduling. However, the scheduling procedure should reduce the traffic load and delay. A parent-dominant decision scheduling with collision free algorithm is proposed to adapt the dynamics of links in a realistic low-power wireless network. In addition, the protocol reduces the scheduling traffic through maintaining distributed lightweight conflict-links tables. We have implemented the protocol in TinyOS and Telosb hardware. The results show that the protocol has robustness to the topology changes and it has significant improvements to reduce the traffic load. Limitations: The experiment has been carried out in an office environment, where 24 Telosb motes were on the 2nd, 3rd, 4th and 5th floors in the building. We have col- lected the schedules of each node every 8 seconds through a single hop or multiple hops. However, other node deployments have not been investigated in the experimental evaluation. Additionally, performance comparisons with more scheduling protocols need to be further investigated. Summary of Contributed Papers 43

Paper V Joint Routing and MAC for Critical Trafﬁc in Industrial Wireless Sensor and Actuator Networks

This paper proposes Delay Bounded Routing (DBR) and extends the Priority- MAC to the network layer by means of the joint DBR and the PriorityMAC, designed for routing critical traffic in IWSAN. Exceeding the required delay bound for unpredictable and emergency traffic could lead to system instability, economic and material losses, system failure and, ultimately, a threat to human safety. Guaranteeing the timely delivery of the critical traffic and its prioritization over regular traffic (e.g. non-critical monitoring traffic) is a significant challenge. Therefore, we present the design, implementation and evaluation of DBR and PriorityMAC. DBR and Priority- MAC are implemented in TinyOS and evaluated on a testbed of Telosb motes. The experimental results indicate that our protocol achieves a significant improvement with regards to the end-to-end delivery latency compared to the current industrial standards. Limitations: This paper lacks of a theoretical analysis on multi-hop end-to-end delay performance. Additionally, a comparison with other routing protocols on top of PriorityMAC is not presented. 44 Chapter 4

Conclusions and Future Work

4.1 Conclusions

The emerging technology of industrial wireless sensor and actuator networks has initiated a paradigm shift in industrial automation applications. However, such wireless solutions have been introducing new challenges as well. The overall aim of the research project is to propose solutions to solve vital challenges such as real-time, service differentiation, adaption to link dynamics, security, wireless coexistence and etc. . The objective of this research work has been to design algorithms and protocols to enhance service differentiation for TDMA-based wireless medium access and adapt to link dynamics for scheduling algorithms on top of real-time communications in IWSAN, and to develop a prototype to evaluate and demonstrate potential of proposed solutions. In order to achieve the objective, the thesis proposes a protocol framework for adaptive real-time communication in IWSAN. The protocol framework mainly consists of proposed algorithms and protocols in five contributed articles. Firstly, the solutions in Paper I, Paper II and Paper V have been proposed to enhance service differentiation for TDMA-based wireless medium access in IWSAN. Paper I proposes PriorityMAC, a Priority enhanced Medium Access Control protocol. To the best of our knowledge, this is the first priority-enhanced MAC protocol compatible with industrial standards for IWSAN. Paper V proposes Delay Bounded Routing to extend the PriorityMAC to network layer by joint DBR and the Priority- MAC, designed for routing critical traffic in IWSAN. The experimental results indicate that the solutions efficiently handles different traffic categories with different latency requirements, thereby achieving a significant improvement in the delivery latency compared with the current industrial standards. The current industrial standards provide a limited service differentiation for different traffic categories in medium access control. Based on the schemes in these standards, Paper II proposes a CCA-Embedded TDMA scheme that improves the transmission efficiency for burst acyclic traffic. The analytical, simulation and experimental results show that the

45 46 Conclusions and Future Work

solution can efficiently increase the throughput and reduce the expected delay compared with WirelessHART. Secondly, scheduling algorithms on top of real-time communications have been proposed to adapt to link dynamics in IWSAN in Papers III and IV. Paper III presents SAS-TDMA, a Source-Aware Scheduling algorithm which uses a cross-layer solution to adapt to link dynamics in realistic environment. Paper IV further proposes a novel distributed scheduling protocol for data gathering transmission. It is also able to adapt the dynamics of links in a realistic low-power wireless network. The purpose to adapt to link dynamics for scheduling algorithms significantly differentiates the contributions in the two papers from ones in most of state-of-the-art scheduling algorithms articles. Numerical results show that the solutions achieve significant improvements for realistic dynamic wireless sensor networks when compared to existing scheduling algorithms with the aim to minimize latency for real-time communication. Lastly, a test-bed consisting of about 100 Telosb and Micaz devices has been built up with a software stack. I have implemented the above proposed protocol framework in the stack and evaluated the proposed solutions in the test-bed.

4.2 Future work

Future work will deal with some speciﬁc improvements for the proposed protocols and algorithms. Firstly, collisions among highest-priority packets from nodes shared a medium can be solved in a smarter manner for PriorityMAC. Furthermore, SAS-TMDA and DBR need further evaluations in real platforms and comparisons with other routing protocols such as ROLL RPL. Thirdly, the proposed protocol framework combining PriorityMAC, adaptive routing algorithms and cross-layer scheduling algorithms will be further evaluated in industrial harsh environment. Furthermore, new challenges, such as lack of compatibility among the industrial standards for industrial automation applications, require novel ideas and technologies. For example, the dynamic spectrum access concept promoted by technologies such as cognitive radio systems might promote next-generation wireless technologies to solve the practical issues in a much more satisfactory manner. The resource- constrained IP-based technologies can also promote convergence solutions from a technique standpoint. Last but not least, the low-power and resource-constrained wireless technology combining Internet of Things will bring more new opportunities and challenges for the future research. Bibliography

[1] M. Palattella, N. Accettura, X. Vilajosana, T. Watteyne, L. Grieco, G. Boggia, and M. Dohler, “Standardized protocol stack for the internet of (important) things,” IEEE Commun. Surveys Tuts. , vol. 15, no. 3, pp. 1389–1406, 2013. [2] IEEE Standard for Information technology– Local and metropolitan area networks– Specific requirements– Part 15.4: Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low Rate Wireless Personal Area Networks (WPANs), IEEE Std 802.15.4, 2006. [3] J. Akerberg,˚ “On safe and secure communication in process automation,” Ph.D. dissertation, Malardalen¨ University Press, Nov. 2011. [4] “National Pooling of Resources for Industrial Process Automation.” [Online]. Available: http://www.processitinnovations.se/Sve/Nyheter/Filer/national pooling 120619.pdf [5] Industrial communication networks – Wireless communication network and communication profiles – WirelessHART TM , International Electrotechnical Commission (IEC) Std. 62 591, 2010. [6] Wireless Systems for Industrial Automation: Process Control and Related Applica- tions, ISA100.11a Standard, 2009. [7] K. Pister, P. Thubert, S. Dwars, and T. Phinney, Industrial Routing Requirements in Low-Power and Lossy Networks , The Internet Engineering Task Force (IETF) RFC 5673, Oct. 2009. [8] T. Sauter, “The three generations of field-level networks – evolution and compatibility issues,” IEEE Trans. Ind. Electron. , vol. 57, no. 11, pp. 3585–3595, Nov. 2010. [9] D. Chen, M. Nixon, and A. Mok, WirelessHART TM – Real-Time Mesh Network for Industrial Automation. New York, NY, USA: Springer Science+Business Media, LLC, 2010. [10] J. Akerberg,˚ M. Gidlund, T. Lennvall, J. Neander, and M. Bj orkman,¨ “Efficient integration of secure and safety critical industrial wireless sensor networks,” EURASIP Journal on Wireless Communications and Networking , Nov. 2011.

47 48 BIBLIOGRAPHY

[11] K. Srinivasan, M. A. Kazandjieva, S. Agarwal, and P. Levis, “The β-factor: Measuring wireless link burstiness,” in Proc. 6th ACM Conf. Embedded Network Sensor Systems (SenSys), 2008, pp. 29–42.

[12] K. Srinivasan, P. Dutta, A. Tavakoli, and P. Levis, “An empirical study of low- power wireless,” ACM Trans. Sen. Netw. , vol. 6, no. 2, pp. 16:1–16:49, Feb. 2010.

[13] D. Stanislowski, X. Vilajosana, Q. Wang, T. Watteyne, and K. Pister, “Adap- tive Synchronization in IEEE 802.15.4e Networks,” IEEE Trans. Ind. Informat. , vol. 10, no. 1, pp. 795–802, 2014.

[14] D. Gay, P. Levis, and D. Culler, “Software design patterns for TinyOS,” ACM Trans. Embed. Comput. Syst. , vol. 6, no. 4, Sept. 2007.

[15] TinyOS. [Online]. Available: http://www.tinyos.net/

[16] Contiki. [Online]. Available: http://www.contiki-os.org/

[17] T. Watteyne, X. Vilajosana, B. Kerkez, F. Chraim, K. Weekly, Q. Wang, S. Glaser, and K. Pister, “OpenWSN: a standards-based low-power wireless development environment,” Trans. Emerging Telecommunications Technologies , vol. 23, no. 5, pp. 480–493, 2012.

[18] P. Levis, N. Lee, M. Welsh, and D. Culler, “TOSSIM: accurate and scalable simulation of entire TinyOS applications,” in Proc. 1st Int. Conf. Embedded Net- worked Sensor Systems (SenSys), 2003, pp. 126–137.

[19] IEEE Standard for Local and metropolitan area networks – Part 15.4: Low-Rate Wire- less Personal Area Networks (LR-WPANs) – Amendment 1: MAC sublayer , IEEE Std. 802.15.4eTM , April 2012.

[20] E. Kim, D. Kaspar, and J. Vasseur, IPv6 over Low-Power Wireless Personal Area Networks (6LoWPANs): Overview, Assumptions, Problem Statement, and Goals , In- ternet Engineering Task Force (IETF), RFC 6568, April 2012.

[21] T. Winter, P. Thubert, A. Brandt, J. Hui, R. Kelsey, P. Levis, K. Pister, R. Struik, J. P. Vasseur, and R. Alexander, RPL: IPv6 Routing Protocol for Low-Power and Lossy Networks, Internet Engineering Task Force, RFC 6550, Mar. 2012.

[22] Constrained RESTful Environments (CORE) , “IETF Working Group.” [Online]. Available: https://datatracker.ietf.org/wg/core/charter/

[23] J.-P. Vasseur and A. Dunkels, Interconnecting Smart Objects with IP: The Next Internet. San Francisco, CA, USA: Morgan Kaufmann Publishers Inc., 2010.

[24] B. SIG, “Bluetooth speciﬁcation version 4.0 [vol 0],” June 2010. [Online]. Available: http://www.bluetooth.org

[25] S. Aust, R. Prasad, and I. Niemegeers, “IEEE 802.11ah: Advantages in standards and further challenges for sub 1 GHz Wi-Fi,” in Proc. IEEE Int. Conf. Communications (ICC), June 2012, pp. 6885–6889. BIBLIOGRAPHY 49

[26] W. Sun, M. Choi, and S. Choi, “IEEE 802.11ah: A long range 802.11 WLAN at sub 1 GHz,” Journal of ICT Standardization , vol. 1, pp. 83–108, May 2013.

[27] A. Willig, “Recent and emerging topics in wireless industrial communications: A selection,” IEEE Trans. Ind. Informat. , vol. 4, no. 2, pp. 102–124, May 2008.

[28] S. Shin, T. Kwon, G.-Y. Jo, Y. Park, and H. Rhy, “An experimental study of hierarchical intrusion detection for wireless industrial sensor networks,” IEEE Trans. Ind. Informat. , vol. 6, no. 4, pp. 744–757, Nov. 2010.

[29] T. Chiwewe and G. Hancke, “A distributed topology control technique for low interference and energy efﬁciency in wireless sensor networks,” IEEE Trans. Ind. Informat., vol. 8, no. 1, pp. 11–19, Feb. 2012.

[30] L. Palopoli, R. Passerone, and T. Rizano, “Scalable ofﬂine optimization of industrial wireless sensor networks,” IEEE Trans. Ind. Informat. , vol. 7, no. 2, pp. 328–339, May 2011.

[31] G. Anastasi, M. Conti, and M. Di Francesco, “Extending the lifetime of wireless sensor networks through adaptive sleep,” IEEE Trans. Ind. Informat. , vol. 5, no. 3, pp. 351–365, Aug. 2009.

[32] V. Gungor and G. Hancke, “Industrial wireless sensor networks: challenges, design principles, and technical approaches,” IEEE Trans. Ind. Electron. , vol. 56, no. 10, pp. 4258–4265, Oct. 2009.

[33] K. Al Agha, M.-H. Bertin, T. Dang, A. Guitton, P. Minet, T. Val, and J.-B. Viol- let, “Which wireless technology for industrial wireless sensor networks? The development of OCARI technology,” IEEE Trans. Ind. Electron. , vol. 56, no. 10, pp. 4266–4278, Oct. 2009.

[34] J. Akerberg,˚ M. Gidlund, and M. Bj orkman,¨ “Future research challenges in wireless sensor and actuator networks targeting industrial automation,” in Proc. 9th IEEE Int. Conf. Industrial Informatics (INDIN) , July 2011, pp. 410–415.

[35] K.-S. Low, W. Win, and M.-J. Er, “Wireless sensor networks for industrial environments,” in Proc. Int. Conf. Computational Intelligence for Modelling, Control and Automation, and Int. Conf. Intelligent Agents, Web Technologies and Internet Commerce , vol. 2, Nov. 2005, pp. 271–276.

[36] D. Yang, Y. Xu, and M. Gidlund, “Coexistence of IEEE802.15.4 based networks: A survey,” in Proc. IECON 2010 - 36th Annual Conf. IEEE Industrial Electronics Society, Nov. 2010, pp. 2107–2113.

[37] ZigBee Alliance. [Online]. Available: https://www.zigbee.org/

[38] C. Caione, D. Brunelli, and L. Benini, “Distributed compressive sampling for lifetime optimization in dense wireless sensor networks,” IEEE Trans. Ind. In- format., vol. 8, no. 1, pp. 30–40, Feb. 2012. 50 BIBLIOGRAPHY

[39] P. T. A. Quang and D.-S. Kim, “Enhancing real-time delivery of gradient routing for industrial wireless sensor networks,” IEEE Trans. Ind. Informat. , vol. 8, no. 1, pp. 61–68, Feb. 2012.

[40] P. Suriyachai, U. Roedig, and A. Scott, “A survey of MAC protocols for mission-critical applications in wireless sensor networks,” IEEE Commun. Sur- veys Tuts. , vol. 14, no. 2, pp. 240–264, 2012.

[41] Industrial communication networks – Fieldbus speciﬁcations – WIA-PA communication network and communication proﬁle , International Electrotechnical Commis- sion (IEC) Std. 62 061, 2011.

[42] K. Pister and L. Doherty, “TSMP: Time synchronized mesh protocol,” in Proc. IASTED Int. Symp. Distributed Sensor Networks (DSN08), Nov. 2008, pp. 391– 398.

[43] J. Song, S. Han, A. Mok, D. Chen, M. Lucas, M. Nixon, and W. Pratt, “Wire- lessHART: Applying wireless technology in real-time industrial process control,” in Proc. 2008 IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS ’08) , 2008, pp. 377–386.

[44] W. Liang, S. Liu, Y. Yang, and S. Li, “Research of adaptive frequency hopping technology in WIA-PA industrial wireless network,” in Advances in Wireless Sensor Networks, ser. Communications in Computer and Information Science, 2013, vol. 334, pp. 248–262.

[45] H. Karl and A. Willig, Protocols and Architectures for Wireless Sensor Networks . John Wiley & Sons Ltd., April 2005.

[46] T. S. Rappaport, Wireless Communications: Principles and Practice , 2nd ed. Pren- tice Hall, 2002.

[47] A. Bachir, M. Dohler, T. Watteyne, and K. Leung, “MAC essentials for wireless sensor networks,” IEEE Commun. Surveys Tuts. , vol. 12, no. 2, pp. 222–248, 2010.

[48] W. Ye, J. Heidemann, and D. Estrin, “An energy-efﬁcient MAC protocol for wireless sensor networks,” in Proc. INFOCOM , vol. 3, 2002, pp. 1567 – 1576.

[49] I. Demirkol, C. Ersoy, and F. Alagoz, “MAC protocols for wireless sensor networks: a survey,” IEEE Communications Magazine, vol. 44, no. 4, pp. 115–121, Apr. 2006.

[50] J. Polastre, J. Hill, and D. Culler, “Versatile low power media access for wireless sensor networks,” in Proc. 2nd Int. Conf. Embedded Networked Sensor Systems (SenSys), 2004, pp. 95–107.

[51] A. El-Hoiydi and J.-D. Decotignie, “WiseMAC: an ultra low power MAC protocol for the downlink of infrastructure wireless sensor networks,” in Proc. 9th Int. Symp. Computers and Communications (ISCC 2004) - Volume 02 , 2004, pp. 244–251. BIBLIOGRAPHY 51

[52] A. El-Hoiydi, “Spatial tdma and csma with preamble sampling for low power ad hoc wireless sensor networks,” in Proc. 7th Int. Symp. Computers and Com- munications (ISCC 2002), 2002, pp. 685–692. [53] I. Rhee, A. Warrier, M. Aia, J. Min, and M. Sichitiu, “Z-MAC: A hybrid MAC for wireless sensor networks,” IEEE/ACM Trans. Netw. , vol. 16, no. 3, pp. 511 –524, June 2008. [54] S. Munir, S. Lin, E. Hoque, S. M. S. Nirjon, J. A. Stankovic, and K. Whitehouse, “Addressing Burstiness for Reliable Communication and Latency Bound Gen- eration in Wireless Sensor Networks,” in Proc. 9th ACM/IEEE Int. Conf. Infor- mation Processing in Sensor Networks (IPSN ’10) , 2010, pp. 303–314. [55] P. Suriyachai, J. Brown, and U. Roedig, “Time-critical data delivery in wireless sensor networks,” in Proc. 6th IEEE Int. Conf. Distributed Computing in Sensor Systems (DCOSS’10), 2010, pp. 216–229. [56] F. Cali, M. Conti, and E. Gregori, “IEEE 802.11 protocol: design and performance evaluation of an adaptive backoff mechanism,” IEEE J. Selected Areas in Communications, vol. 18, no. 9, pp. 1774–1786, 2000. [57] IEEE Standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Speciﬁcations, IEEE Std 802.11, 2007. [58] G. Lu, B. Krishnamachari, and C. Raghavendra, “Performance evaluation of the IEEE 802.15.4 MAC for low-rate low-power wireless networks,” in Proc. IEEE Int. Conf. Performance, Computing, and Communications, 2004, pp. 701–706. [59] J. Zheng and M. Lee, “Will IEEE 802.15.4 make ubiquitous networking a re- ality?: A discussion on a potential low power, low bit rate standard,” IEEE Communications Magazine, vol. 42, no. 6, pp. 140–146, 2004. [60] J.-S. Lee, “An experiment on performance study of IEEE 802.15.4 wireless networks,” in Proc. IEEE Conf. Emerging Technologies and Factory Automation (ETFA) , vol. 2, 2005, pp. 451–458. [61] J. Ploennigs, V. Vasyutynskyy, and K. Kabitzsch, “Comparative study of energy-efﬁcient sampling approaches for wireless control networks,” IEEE Trans. Ind. Informat. , vol. 6, no. 3, pp. 416–424, Aug. 2010. [62] G. Anastasi, M. Conti, and M. Di Francesco, “A comprehensive analysis of the MAC unreliability problem in IEEE 802.15.4 wireless sensor networks,” IEEE Trans. Ind. Informat. , vol. 7, no. 1, pp. 52–65, Feb. 2011. [63] S. Ding, P. Zhang, S. Yin, and E. Ding, “An integrated design framework of fault-tolerant wireless networked control systems for industrial automatic control applications,” IEEE Trans. Ind. Informat. , vol. 9, no. 1, pp. 462–471, Feb. 2013. [64] Y. Y. Chewoo Naa and A. Mishrab, “An optimal GTS scheduling algorithm for time-sensitive transactions in IEEE 802.15.4 networks,” Computer Networks, vol. 52, no. 13, pp. 2543–2557, Sept. 2008. 52 BIBLIOGRAPHY

[65] A. Koub aa,ˆ M. Alves, E. Tovar, and A. Cunha, “An implicit GTS allocation mechanism in IEEE 802.15.4 for time-sensitive wireless sensor networks: the- ory and practice,” Real-Time Syst., vol. 39, no. 1-3, pp. 169–204, Aug 2008. [66] Y.-K. Huang, A.-C. Pang, and H.-N. Hung, “An adaptive GTS allocation scheme for IEEE 802.15.4,” IEEE Trans. Parallel Distrib. Syst. , vol. 19, no. 5, pp. 641–651, May 2008. [67] J. Song, J. dong Ryoo, S. Kim, J. Kim, H. Kim, and P. Mah, “A dynamic GTS allocation algorithm in IEEE 802.15.4 for QoS guaranteed real-time applications,” in Proc. IEEE Int. Symp. Consumer Electronics (ISCE) , June 2007, pp. 1–6. [68] Z. Han zlek´ and P. Jur cˇ´ık, “Energy efficient scheduling for cluster-tree wireless sensor networks with time-bounded data flows: Application to IEEE 802.15.4/ZigBee,” IEEE Trans. Ind. Informat. , vol. 6, no. 3, pp. 438–450, Aug. 2010. [69] S.-E. Yoo, P. K. Chong, D. Kim, Y. Doh, M.-L. Pham, E. Choi, and J. Huh, “Guar- anteeing real-time services for industrial wireless sensor networks with IEEE 802.15.4,” IEEE Trans. Ind. Electron. , vol. 57, no. 11, pp. 3868–3876, Nov. 2010. [70] E. Toscano and L. Bello, “Multichannel superframe scheduling for IEEE 802.15.4 industrial wireless sensor networks,” IEEE Trans. Ind. Informat. , vol. 8, no. 2, pp. 337–350, May 2012. [71] S. Li, S. Ren, Y. Yu, X. Wang, L. Wang, and G. Quan, “Profit and penalty aware scheduling for real-time online services,” IEEE Trans. Ind. Informat. , vol. 8, no. 1, pp. 78–89, Feb. 2012. [72] D. Tse and P. Viswanath, Fundamentals of Wireless Communication . Cambridge University Press, 2005. [73] W. Wang, Y. Wang, X.-Y. Li, W.-Z. Song, and O. Frieder, “Efficient interference- aware TDMA link scheduling for static wireless networks,” in Proc. ACM Mo- biCom, 2006, pp. 262–273. [74] W.-Z. Song, F. Yuan, and R. LaHusen, “Time-optimum packet scheduling for many-to-one routing in wireless sensor networks,” in Proc. 3rd IEEE Int. Conf. Mobile Ad-hoc and Sensor Systems (MASS), Oct. 2006, pp. 81–90. [75] G. Chakraborty, “Genetic algorithm to solve optimum tdma transmission schedule in broadcast packet radio networks,” IEEE Trans. Commun. , vol. 52, no. 5, pp. 765–777, May 2004. [76] S. Gandham, M. Dawande, and R. Prakash, “Link scheduling in sensor networks: distributed edge coloring revisited,” in Proc. IEEE INFOCOM , vol. 4, Mar. 2005, pp. 2492–2501. [77] H. Zhang, P. Soldati, and M. Johansson, “Optimal link scheduling and channel assignment for convergecast in linear WirelessHART networks,” in Proc. 7th Int. Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOPT) , June 2009, pp. 1–8. BIBLIOGRAPHY 53

[78] P. Djukic and S. Valaee, “Delay aware link scheduling for multi-hop TDMA wireless networks,” IEEE/ACM Trans. Netw. , vol. 17, no. 3, pp. 870–883, June 2009.

[79] H. Li, P. Shenoy, and K. Ramamritham, “Scheduling messages with deadlines in multi-hop real-time sensor networks,” in Proc. 11th IEEE Real Time on Em- bedded Technology and Applications Symposium (RTAS ’05) , 2005, pp. 415–425.

[80] S. C. Ergen and P. Varaiya, “TDMA scheduling algorithms for wireless sensor networks,” Wirel. Netw. , vol. 16, no. 4, pp. 985–997, May 2010.

[81] M. Jain, J. I. Choi, T. Kim, D. Bharadia, S. Seth, K. Srinivasan, P. Levis, S. Katti, and P. Sinha, “Practical, real-time, full duplex wireless,” in Proc. 17th Annual Int. Conf. Mobile Computing and Networking (MobiCom ’11), 2011, pp. 301–312.

[82] L. Shi and A. Fapojuwo, “TDMA scheduling with optimized energy efﬁciency and minimum delay in clustered wireless sensor networks,” IEEE Trans. Mobile Computing, vol. 9, no. 7, pp. 927–940, July 2010.

[83] S. Cui, R. Madan, A. Goldsmith, and S. Lall, “Energy-delay tradeoffs for data collection in TDMA-based sensor networks,” in Proc. IEEE Int. Conf. Commun. (ICC 2005), vol. 5, May 2005, pp. 3278–3284.

[84] C. Ngo and V. Li, “Centralized broadcast scheduling in packet radio networks via genetic-ﬁx algorithms,” IEEE Trans. Commun. , vol. 51, no. 9, pp. 1439–1441, Sept. 2003.

[85] J. F. Kurose and K. W. Ross, Computer Networking: A Top-Down Approach , 6th ed. Pearson Education, Limited, 2012.

[86] K. Akkaya and M. Younis, “A survey on routing protocols for wireless sensor networks,” Ad Hoc Networks , vol. 3, no. 3, pp. 325 – 349, 2005.

[87] I. Stojmenovic, Handbook of sensor networks: algorithms and architectures . John Wiley & Sons, Inc., Hoboken, New Jersey, Oct. 2005, ch. Geographic and Energy-Aware Routing in Sensor Networks, pp. 381-416.

[88] Y. Yu, R. Govindan, and D. Estrin, “Geographical and energy aware routing: A recursive data dissemination protocol for wireless sensor networks,” UCLA Computer Science Department Technical Report UCLA/CSD-TR-01- 0023, Tech. Rep., May 2001.

[89] W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “An application- speciﬁc protocol architecture for wireless microsensor networks,” IEEE Trans. Wireless Commun. , vol. 1, no. 4, pp. 660–670, Oct. 2002.

[90] A. Manjeshwar and D. Agrawal, “TEEN: a routing protocol for enhanced ef- ﬁciency in wireless sensor networks,” in Proc. 15th Int. Parallel and Distributed Processing Symposium. , Apr. 2001, pp. 2009–2015. 54 BIBLIOGRAPHY

[91] O. Younis and S. Fahmy, “HEED: a hybrid, energy-efﬁcient, distributed clustering approach for ad hoc sensor networks,” IEEE Trans. Mobile Comput. , vol. 3, no. 4, pp. 366–379, Oct. 2004.

[92] C. Schurgers and M. Srivastava, “Energy efﬁcient routing in wireless sensor networks,” in Military Communications Conference (MILCOM 2001), vol. 1, 2001, pp. 357–361.

[93] F. Ye, G. Zhong, S. Lu, and L. Zhang, “GRAdient Broadcast: A robust data delivery protocol for large scale sensor networks,” Wirel. Netw. , vol. 11, no. 3, pp. 285–298, May 2005.

[94] J. Al-Karaki and A. Kamal, “Routing techniques in wireless sensor networks: a survey,” IEEE Wireless Communications , vol. 11, no. 6, pp. 6–28, Dec 2004.

[95] T. Watteyne, A. Molinaro, M. Richichi, and M. Dohler, “From MANET To IETF ROLL standardization: A paradigm shift in WSN routing protocols,” IEEE Commun. Surveys Tuts. , vol. 13, no. 4, pp. 688–707, 2011.

[96] C. Perkins, E. Belding-Royer, and S. Das, Ad hoc on-demand distance vector (AODV) routing , Internet Engineering Task Force (IETF), RFC 3561, July 2003.

[97] T. Clausen and P. Jacquet, Optimized Link State Routing Protocol (OLSR) , Internet Engineering Task Force (IETF), RFC 3626, 2003.

[98] R. Ogier, F. Templin, and M. Lewis, Topology Dissemination Based on Reverse- Path Forwarding (TBRPF), Internet Engineering Task Force (IETF), RFC 3684, 2004.

[99] D. Johnson, Y. Hu, and D. Maltz, The Dynamic Source Routing Protocol (DSR) for Mobile Ad Hoc Networks for IPv4 , Internet Engineering Task Force (IETF), RFC 4728, 2007.

[100] Mobile Ad-hoc Networks (MANET) , “IETF Working Group.” [Online]. Available: https://datatracker.ietf.org/wg/manet/charter/

[101] A. Woo, T. Tong, and D. Culler, “Taming the underlying challenges of reliable multihop routing in sensor networks,” in Proc. 1st Int. Conf. Embedded Net- worked Sensor Systems (SenSys), 2003, pp. 14–27.

[102] O. Gnawali, R. Fonseca, K. Jamieson, D. Moss, and P. Levis, “Collection tree protocol,” in Proc. 7th ACM Conf. Embedded Networked Sensor Systems (SenSys) , 2009, pp. 1–14.

[103] J. Ko, S. Dawson-Haggerty, O. Gnawali, D. Culler, and A. Terzis, “Evaluat- ing the Performance of RPL and 6LoWPAN in TinyOS,” in Proc. Extending the Internet to Low power and Lossy Networks (IP+SN 2011) , April 2011.

[104] N. Tsiftes, J. Eriksson, and A. Dunkels, “Low-power wireless IPv6 routing with ContikiRPL,” in Proc. 9th ACM/IEEE Int. Conf. Information Processing in Sensor Networks (IPSN ’10), 2010, pp. 406–407. BIBLIOGRAPHY 55

[105] J. Ko, A. Terzis, S. Dawson-Haggerty, D. Culler, J. Hui, and P. Levis, “Con- necting low-power and lossy networks to the internet,” IEEE Communications Magazine, vol. 49, no. 4, pp. 96–101, Apr. 2011.

[106] A. Dunkels, B. Gronvall, and T. Voigt, “Contiki - A lightweight and ﬂexible operating system for tiny networked sensors,” in Proc. 29th Annual IEEE Int. Conf. Local Computer Networks (LCN 04), 2004, pp. 455–462. [107] Y. Li, C. S. Chen, Y.-Q. Song, Z. Wang, and Y. Sun, “Enhancing real-time delivery in wireless sensor networks with two-hop information,” IEEE Trans. Ind. Informat., vol. 5, no. 2, pp. 113 –122, May 2009.

[108] P. T. A. Quang and D.-S. Kim, “Enhancing real-time delivery of gradient routing for industrial wireless sensor networks,” IEEE Trans. Ind. Informat. , vol. 8, no. 1, pp. 61 –68, Feb. 2012.

[109] K. Yu, M. Gidlund, J. Akerberg,˚ and M. Bj orkman,¨ “Reliable real-time routing protocol for industrial wireless sensor and actuator networks,” in Proc. 8th IEEE Conf. Industrial Electronics and Applications (ICIEA), June 2013, pp. 1895– 1901.

[110] P. Levis, N. Patel, D. Culler, and S. Shenker, “Trickle: A self-regulating algorithm for code propagation and maintenance in wireless sensor networks,” in Proc. 1st Symp. Networked Systems Design and Implementation (NSDI’04) , 2004, pp. 15–28.

[111] I. E. Workbench. [Online]. Available: http://www.iar.com/Products/IAR- Embedded-Workbench/

[112] P. Levis, “Experiences from a decade of tinyos development,” in Proc. 10th USENIX Conf. Operating Systems Design and Implementation (OSDI’12), 2012, pp. 207–220. [113] M. Mar oti,´ B. Kusy, G. Simon, and A. L edeczi,´ “The ﬂooding time synchronization protocol,” in Proc. 2nd Int. Conf. Embedded Networked Sensor Systems (SenSys ’04), 2004, pp. 39–49.

[114] M. Durvy, J. Abeill e,´ P. Wetterwald, C. O’Flynn, B. Leverett, E. Gnoske, M. Vi- dales, G. Mulligan, N. Tsiftes, N. Finne, and A. Dunkels, “Making sensor networks IPv6 ready,” in Proc. 6th ACM Conf. Embedded Network Sensor Systems (SenSys), Nov. 2008, pp. 421–422.