Automation & Control Technical

Wireless ad-hoc networks for industrial automation: trends and prospects by Mogens Mathiesen, Gilles Thonet and Niels Aakvaag, ABB Norway

This article provides an overview of recent advances in wireless communication technologies applied to industrial automation.

Ad-hoc networks typically consist of a set of mobile power consumption, and usability has been nodes that join or leave the network with minimal released by the Bluetooth SIG. With the growing hype around wireless personal overhead. Nodes within transmission range area networks (WPANs), Bluetooth was profiled Suitability for industrial automation. Bluetooth communicate directly, whilst communication as a generic short-range ad-hoc networking has some properties that limit its use in industrial over longer distances uses intermediate nodes technology. Large portions of its specifications environments: as relays. Network operation is said to be self- are now part of the standard IEEE 802.15.1 organised and self-healing: should a transmission • Its power consumption is high. It does not (IEEE Std. 802.15.1, 2002). This effort has lost some compare favourably with standards like path fail, alternative paths are automatically momentum due to both fierce competition and ZigBee. established. inherent limitations of Bluetooth. • Piconets may only contain up to seven active Fig. 1 shows an example of an industrial ad Technical overview: Bluetooth operates in the slave nodes. hoc network. The normal path between 2,4 GHz license-free band using frequency • Sleeping nodes have very long wake-up station 1 and device 4 is represented by a plain hopping spread spectrum (FHSS). The band times, making power optimisation a tough line. If any part of this route fails, data can be is divided into 79 separate channels hopping challenge. routed via the dotted path instead. at a nominal rate of 1’600 hops per second. WLAN Technological overview The main advantage of frequency hopping is increased resilience to adverse Wireless local area network (WLAN) is divided During the short history of wireless ad-hoc interference. Bluetooth guarantees low latency into three main sub-standards, namely a, b, networking, several vendors have developed for active nodes but wake up for sleeping nodes proprietary solutions. There are two fundamental is about three seconds. The network consists of problems using proprietary networks: a master and up to seven slave nodes interoperability across vendors and availability of in a star topology (Fig. 2. Bluetooth nodes may chips and systems. Therefore, many technology belong to more than one network and form so providers and market leaders are now adopting called scatternets (Fig. 3). wireless standards. The ones (existing or emerging) relevant for industrial settings are discussed in the Technology roadmap. A three-year plan next subsections. including enhancements to performance,

Fig. 2. Bluetooth piconet (source: Bluetooth SIG).

Fig. 1. Example of an industrial ad-hoc network. Fig. 3. Bluetooth scatternet (source: Bluetooth SIG).

26 March 2007 - EngineerIT and g, complemented by a set of additional specifications for enhanced security, performance, etc. WLAN is infrastructure-based, meaning that each device connects as a client to fixed access points (APs).

WLAN is intended to replace wired Ethernet connections. Typical usage patterns are mobile connections in office or community-like local networks.

Technical overview. The three main sub standards offer distinct features:

• 802.11b operates in the 2.4 GHz band, ensures Fig. 4. ZigBee network topologies (source: ZigBee Alliance). a range of about 100 m, and offers data rates of up to 11 Mbps (IEEE Std 802.11b, 1999). • 802.11a operates in the 5 GHz band, thus offering shorter range but at a data rate of 54 Mbps to 108 Mbps (IEEE Std 802.11a, 1999). • 802.11g operates in the 2.4 GHz band while communicating at up to 54 Mbps. It is backward-compatible with 802.11b (IEEE Std. 802.11g, 2003).

These are all infrastructure-based, but they also include an ad-hoc mode that allows for one- transmissions between APs and mobile devices.

Technology roadmap: A new version, the 802.11s is being defined. It will enabling mesh connectivity in the 802.11 family with ad-hoc networks of up to 32 nodes. It will bring multihopping capability to the base standards.

Suitability for industrial automation: Due to Fig. 5. Comparison between current wireless standards. higher power requirements WLAN is not suitable for communication between small autonomous

Parameter Bluetooth WLAN ZigBee industrial devices. Range 10 m 100 m 30-100 m ZigBee Associated standard IEEE 802.15.1 IEEE 802.11a/b/g IEEE 802.15.4 ZigBee is a new international standard for Frequency bands 2,4 GHz 2,4 GHz or 5 GHz 868 MHz, 915 MHz, 2.4 GHz network connectivity based on the IEEE Physical layer FHSS DSSS or OFDM DSSS 802.15.4 specification (IEEE Std. 802.15.4, Maximal gross data 1 Mbps 5.5/11/54 Mbps (depending 20/40/250 kbps (depending 2003). The ZigBee Alliance is in charge of further rate on substandard) on channel) development, standardisation, and worldwide Average RF power 1/2.5/100 mW 40-800 mW (depending on 200-500 µW (depending on marketing. (depending on range) substandard) duty cycle) ZigBee specifically addresses the energy usage. Battery life 1-7 days 0,5-5 days 100-1’000+ days A device should be able to run for many years Maximal range 10-100 m 20-100 m 30-100 m on the same batteries. This is achieved by Network topology Star, piconet, scatternet Star Star, mesh, cluster-tree having a very low duty cycle and by avoiding Maximal 7 slaves in a piconet 32 APs 65’000 tight synchronisation. Number of Nodes Technical overview: ZigBee relies on the Node acquisition 3 s 2 s 30 ms time IEEE 802.15.4 physical layer and operates in various unlicenced bands worldwide: 2,4 GHz, Node wake-up time 3 s 1 s 15 ms 915 MHz, and 868 MHz. Raw data rates System resources 250+ kB 1+ MB 4-20 kB of 250 kbps can be achieved at 2,4 GHz Expected main use Cable replacement Wireless Ethernet, mobile Monitoring and control, (16 channels). Transmission range is in the region office autonomous devices from 30 to 100 m, depending on power output Table 1 Comparison between current wireless standards. and environmental characteristics.

28 March 2007 - EngineerIT ZigBee defines the network, security, and application framework profile layers. The network layer supports three networking topologies: star, mesh, and cluster tree as shown in Fig. 4.

Suitability for industrial automation: ZigBee has been developed especially for ad-hoc networking applications involving low duty cycles and low data rates and is therefore well suited.

Comparison

Comparisons between the different current Fig. 6. Wireless open-loop systems. wireless standards are shown in Fig. 5 and in Table 1. As they address different applications, they are usually not direct competitors.

Only ZigBee has been specifically designed for ad-hoc networking using a large number of nodes. On the other hand, some industrial applications such as the mobile maintenance scenario described ‘Section Error! Reference source not found’ require higher data rates and therefore justify the use of WLAN.

Open-loop applications

Open-loop applications are characterised by having no wireless connection in the feedback loop of the control system, as shown in Fig. 6. Fig. 7. Mobile maintenance.

A real plant is controlled via a number of actuators. The operation of the plant gives rise to some measurable output that is sensed. The wireless links are associated with processing or operations that are not part of the classical feedback path of the control system. Processing of off-line data is typically non real-time and is used to deduce some additional information subsequently fed either to the control system or embedded directly into the actual plant.

Mobile maintenance

In a process control environment, service personnel frequently need to communicate with field devices, either for test, calibration, or fault tracking. Traditionally, staff connects Fig. 8. Process control. to the device with a cable and performs the appropriate physical maintenance. Process control The set of requirements will obviously vary greatly depending on configuration and sensor Adding wireless capabilities to field assets Another application of considerable interest is type. greatly simplifies plant operations. First, the getting access to field instruments that are not actual connection to the device is no longer typically connected and whose data do not Ad-hoc benchmarking physical. Second, by being in radio contact with necessarily constitute an integral part of the the devices it is easy to provide the user with main control loop. One example is historical and A fascinating new topic in process control is location-sensitive information (Fig. 7). real time temperature data from within rotating ad hoc benchmarking (Fig. 9). The operation machinery. This location is difficult to access with of a plant is suspected to be suboptimal. requirements may be considerable wired equipment. Data transfer rates are typically Removable sensors are placed at central if historical or maintenance data is transferred low and have relaxed latency requirements. locations in the process. over the air interface. However, there will be no restriction on latency and very relaxed power A schematic representation of using ad- Once in place, they establish an ad-hoc constraints, making this an ideal application for hoc wireless networking in a process control and route their measured wireless ad-hoc networks. environment is shown in Fig. 8. values to some aggregation point for off-line

EngineerIT - March 2007 29 or functionality, but also have their own particularities. In order to obtain an efficient implementation of an ad-hoc network in an open loop environment, a thorough analysis of the application is required.

Closed-loop applications

Extending the reach to the control loop itself is the next natural step. Distributed control with feedback loops closed over wireless links is an emerging research topic that is attracting growing attention. Actual deployment for complex control applications is not expected Fig. 9. Ad-hoc benchmarking. for some years. Most contributions to the field to date have been led by universities (Liu, and Goldsmith, 2003; Liu, and Goldsmith, 2004; Ploplys, et al., 2004; Sinopoli, et al., 2004).

Fig. 11 shows a very general closed-loop scenario in which both sensing and actuation are performed through wireless links. Note that actuators may be wired while keeping the wireless sensing loop.

Fig. 10. Mobility. Trading off communication parameters

Building a distributed control system over a wireless sensor/actuator network is not a straightforward task. To keep control systems running smoothly, data transmission over the industrial network should be timely, reliable, and accurate (Liu and Goldsmith, 2003). This becomes challenging with wireless links since they introduce random delays and packet losses due to interference, signal attenuation, Fig. 11. Closed-loop wireless control system. and multipath transmissions.

In this setting, achieving closed-loop control processing. The beauty of this approach is that overall performance of the plant. Fig. 10 shows over wireless links can be rephrased the the system is easy to install, self-configures, and a typical example of this scenario. following way: Designing a decentralised can be reused in different locations. wireless ad-hoc network that minimises the One example of this application is in the pulp impact of communication faults on the control Once a new and optimised control strategy has and paper industry. Sensors are added to the system. been developed, it may be implemented into drying process of the paper mass, measuring the central controller or into individual pieces of humidity and temperature along the chain. Liu and Goldsmith (2003) claim that three equipment. Data is gathered, possibly being relayed from communication parameters need particular sensor to sensor, to the controller. attention from the control perspective: data Mobility rate, latency, and packet loss. Since these are Towards new open-loop applications Consider a processing plant with moving units competing objectives, tradeoffs are required placed on a conveyor belt. As units move From the above discussion it is clear that when designing the communication network. they cannot easily be connected to a fixed there is great potential for using wireless ad- Liu and Goldsmith (2003) conclude that infrastructure. Nevertheless, mobile sensors can hoc networks in industrial automation. The fundamental changes in the link layer design measure physical parameters of interest to the applications share some basic requirements are required for wireless closed-loop control.

30 March 2007 - EngineerIT Ideally, joint optimisation is the best approach, control. The so-called wireless proximity sensor of industrial automation. Although hurdles are with communication and control variables (Apneseth, et al., 2002) boasts a proprietary still to be cleared, many innovative companies optimised simultaneously. that ensures reliable around the world are already paving the way delivery of messages within the short time for fully decentralised, highly adaptive and Process control frames required by current programmable cost-effective automation networks. Wireless closed-loop control is ideally suited to logic control systems. References improve the way processes are automated. Towards real distributed wireless control Simple applications utilising wireless sensor data [1] Apneseth, C., D. Dzung, S. Kjesbu, G. Scheible and directly in the controller are already appearing. Distributed wireless operation between the W. Zimmermann (2002). Wireless – Introducing Wireless Proximity Switches. ABB Review, Vol. 4, pp. controller and sensors/actuators is still in its For instance, Ember designed a mesh network 42–49, www.abb.com/technology. infancy. Initiatives are starting to spread both in to support the water treatment process (Tuck [2] Liu, X. and A.J. Goldsmith (2003). Wireless process control and discrete automation. These and Burgess, 2003). The goal of the network was Communication Tradeoffs in Distributed Control. are usually characterised by being small-scale, to connect turbidity meters in the pipe gallery In: Proc. of 42nd IEEE Conference on Decision and single-hop, and with simple control functionality. back to the control system. Control, pp. 688–694, Maui, Hawaii. In order to achieve complex control in a real [3] Liu, X. and A.J. Goldsmith (2004). Wireless Medium Similar sensing applications in other closed-loop fashion, several research directions Access Control in Networked Control Systems. In: environments with high reliability needs are also should be prioritised: Proc. of American Control Conference, Boston, feasible. For instance, the oil and gas industry USA. • Reliable communication protocols able to is continuously calling for cost reductions since [4] Ploplys, N.J., P.A. Kawka and A.G. Alleyne (2004). route data across large-scale sensor and exploration becomes increasingly expensive. Closed-Loop Control over Wireless Networks. IEEE actuator networks. Control Systems Magazine, Vol. 6, pp. 58–71. Savings in network maintenance, cabling cost • Efficient power conservation schemes that fit [5] Sinopoli, B., L. Schenato, M. Franceschetti, K. Poolla, and weight can bring significant advantages energy-constrained environments. M.I. Jordan and S. Sastry (2004). Kalman Filtering to offshore installations that may increase • Redesign of jointly optimised communication with Intermittent Observations. In: IEEE Transactions their lifetime. Large networks with complex and control algorithms to guarantee a on Automatic Control, Vol. 49, No. 9, pp. 1453 sensing and actuating interactions with highly 1464. smooth migration towards fully wireless critical control systems will require robust and [6] Tuck, A. and C. Burgess (2003). Mesh Meets the automation infrastructures. jointly optimised control and communication Need – Wireless for Industrial Control. ISA Technical algorithms. The authors believe that the successful Information and Communities, www.isa.org. completion of these steps will open up new [7] IEEE Std 802.11a (1999). Wireless LAN Medium Production lines avenues for radically improved and cost- Access Control (MAC) and Physical Layer (PHY) Specifications: High-speed Physical Layer in the 5 Production lines can greatly benefit from new effective industrial automation. GHz Band. wireless ad-hoc networking technologies. Conclusions [8] IEEE Std 802.11b (1999). Wireless LAN Medium Closing control loops over wireless links to sensors Access Control (MAC) and Physical Layer (PHY) and/or actuators is the natural evolution towards This survey has introduced wireless ad-hoc Specifications: High-speed Physical Layer Extension networks and their applications in industrial more distributed automation architectures. in the 2.4 GHz Band. automation. The paper has also made the Mobility is typically not a critical issue for [9] IEEE Std 802.11g (2003). Wireless LAN Medium distinction between open-loop and closed-loop factories, and when it is, movement patterns Access Control (MAC) and Physical Layer (PHY) applications, and presented some example Specifications – Amendment 4: Further High Data correspond to a prior scheduling. scenarios. Rate Extension in the 2.4 GHz Band. High transmission speed is usually not required Due to the fact that open-loop applications [10] IEEE Std 802.15.1 (2002). Wireless Medium Access since factory communications mostly carry Control (MAC) and Physical Layer (PHY) Specifications deliver data that is less process-critical, these small amounts of data, often limited to binary for Wireless Personal Area Networks (WPANs scenarios have already started to appear inputs/outputs. Multihop routing is thus usually [11] IEEE Std 802.15.4 (2003). Wireless Medium in industry. A quick rollout of closed-loop an achievable task in these settings. Access Control (MAC) and Physical Layer (PHY) applications is not expected. It is likely however Specifications for Low-rate Wireless Personal Area At the same time, reliability and energy that some of these scenarios will emerge in Networks (LR-WPANs). conservation are two important factors for slow process environments. For faster process Contact Ntaga Mojapelo, ABB, manufacturing automation. scenarios, more research is needed. Tel (011) 236-7000, Recently, ABB made significant progress in The authors believe that wireless ad-hoc networks [email protected]; addressing these two issues for closed-loop have a huge potential for changing the face [email protected] ☐

EngineerIT - March 2007 31