International Journal of Information and Electronics Engineering, Vol. 2, No. 3, May 2012

EDCA Delay Analysis of Spatial in IEEE 802.11-Based Sensor and Actuator Networks

Mohsen Maadani, Member, IACSIT, Seyed Ahmad Motamedi, and Mohammadreza Soltani

studied in the literature since the release of this amendment Abstract—Low packet delay and packet loss probability are [4]-[9]. Final version of IEEE 802.11n amendment [2] has the two most contradicting requirements of wireless sensor and recently been released, aiming to increase the throughput by actuator networks (WSAN). In this paper, saturation delay utilizing Multiple Input Multiple Output (MIMO) technology characteristics of Multiple Input Multiple Output and some other modifications. Using a feedback from (MIMO)-enabled IEEE 802.11 technology utilizing Enhanced Distributed Channel Access (EDCA) at receiver to the transmitter, the Space-Time Code (STC) and (MAC) and a simple spatial-multiplexing scheme in PHYsical pre-coder are adopted to the channel situations with the target (PHY) layer has been thoroughly analyzed. Results show that of maximizing data-rate while maintaining an acceptable due to the low Signal to Noise Ratio (SNR) in WSAN packet error rate (PER). Ref. [4] has studied saturation delay environment, simple schemes with low spectral performance of EDCA mechanism in Single Input Single efficiency have better delay and reliability performance. Output (SISO) configuration and compared it with a simple Otherwise, the delay will be highly sensitive to SNR, packet payload, number of competing nodes. Spatial Multiplexing (SM) MIMO scheme. It is shown that SM outperforms SISO in both packet delay and PER. In this Index Terms—IEEE 802.11, maximum likelihood receiver, paper, delay and reliability performance behavior of utilizing packet delay, spatial multiplexing. SM in MIMO-enabled WLAN stations for WSAN’s has been thoroughly studied. Due to random nature of packet transmission initiations, which is highly dependent on the I. INTRODUCTION network topology and application, carrying out a theoretical Due to the numerous advantages of wireless technology, it analysis without making several simplifying assumptions is has been used in various consumer devices. Recently, there not an easy task [5]. As an alternative, our choice has been has been much attention in making some parts of wireless trying to assess SM scheme through simulation. This sensor and actuator networks (WSAN), wireless. Being approach sometimes provides better results than some reliable and real-time are the two most challenging oversimplified analytical approaches. requirements of these networks, which contradict with the This paper is organized as follows: Section II reviews the nature of wireless communication medium. Enhancing related work in IEEE 802.11 assessment. Section III presents current wireless technologies and standards towards the simulation scenario and results. The paper is concluded in industrial needs, and if not possible, designing new protocols section IV. is a developing research field [1]. Wireless Local Area Network (WLAN) based on IEEE 802.11 standard [2] is one the most common and popular II. RELATED WORK technologies which has been used as the dominant Ref. [4] has compared SISO and SM MIMO schemes from technology in local wireless networks. One of the problems average and maximum packet delay perspective for 1Mbps of using this standard for industrial applications is the data rate. It shows that SM outperforms SISO from delay, non-deterministic behavior of Carrier Sense Multiple PER and Bit Error Rate (BER) aspects, at the cost of 2 Access/Collision Avoidance (CSMA/CA) mechanism used transceivers per node and a little decoding complication. Ref. in its Medium Access Control (MAC) layer. The MAC [5] has analyzed statistical distribution of response time in amendment IEEE 802.11e was introduced to improve the IEEE 802.11g/e configuration. It presents a computational packet delay for certain applications through traffic model for transmission delays when low-to-medium prioritization. It defines 4 Access Categories (AC), which in concurrent traffic is taken into account. It is shown that for the order of priority are: Voice, Video, Best Effort and low network load (below 20%), response times are generally Background. Voice applications can tolerate latency of about bounded. For higher traffic load (up to 40%), 150 ms and packet loss of up to 1%, whereas latency in some quasi-deterministic and bounded latencies are achieved for WSAN’s must be as low as 10 ms, and much higher selected high-priority messages using EDCA mechanism. reliability is required. Analysis and enhancement of IEEE Ref. [6] using simulation, has analyzed the EDCA behavior 802.11e delay property for real-time applications has been for real-time industrial communication. A Markov chain complicated analytical model for the throughput and access Manuscript received February 26, 2012; revised April 21, 2012. delay of IEEE 802.11 EDCA mechanism under saturation Mohsen Maadani, Seyed Ahmad Motamedi and Mohammadreza Soltani are with the Department of Electrical Engineering, Amirkabir University of condition has been proposed in [7]. References [4] and [8] Technology (AUT), Tehran, Iran (e-mail: {maadani, motamedi, have studied the effect of traffic prioritization in aut_mohammad}@aut.ac.ir). IEEE802.11e for real-time applications using simulation. To

318 International Journal of Information and Electronics Engineering, Vol. 2, No. 3, May 2012 the author's knowledge, no other analysis has been done Real-Time traffic is mapped to highest priority (AC0) and regarding the packet delay and reliability of using spatial its characteristics have been analyzed. multiplexing in wireless sensor and actuator networks based Mesh topology with single-hop transmissions has been on MIMO-Enabled WLAN stations. chosen; each node can send packet to all other nodes. Such configuration is typical in sensors and actuators connected to TABLE I: IEEE802.11E SIMULATION SCENARIO the same instrument in a factory (e.g. control loops etc.) Parameter Value Simulation has been done using MATLAB 1 & 2 Mbps (DBPSK and DQPSK communications toolbox. Data Rate ) MIMO:SM Average Delay(n=15) 300 Control Rate 1 Mbps (DBPSK Modulation) 1Mbps (PL=8 Bytes) PHY Header 24 Bytes 1Mbps (PL=32 Bytes) MAC Header 28 Bytes 250 2Mbps (PL=8 Bytes) ACK Packet size 14 Bytes 2Mbps (PL=32 Bytes)

Slot Time 20 µs 200 SIFS 10 µs AIFSN 2 150 CWmin [0 1] 7 15 CWmax [0 1] 15 31 Delay (mS) TXop Disabled 100 Retry Limit Disabled (Unlimited) 50

0 III. SIMULATION RESULTS 2 4 6 8 10 12 14 16 18 20 SNR (dB) A. Simulation Scenario (a) MIMO:SM Average Delay (n=15) IEEE 802.11b with default EDCA parameters listed in 900 1Mbps (SNR=6dB) Table I [2] has been chosen as PHY and MAC layers, except 800 1Mbps (SNR=20dB) 2Mbps (SNR=6dB) for AIFSN (Arbitration Interframe Space for access category 700 2Mbps (SNR=20dB)

N) which is considered the same for all AC’s; i.e. its 600 differentiation effect is not in the scope of this paper. SIFS 500 stands for Short Interframe Space, CWmin and CWmax are 400 minimum Contention Window and maximum Contention Delay (mS) Window respectively. TXop is the Transmission Opportunity 300 [1]. 200 PHY layer preamble and header and packet 100 acknowledgement (ACK) are sent with control rate (1Mbps). 0 0 50 100 150 200 250 300 MAC layer header and payload are sent with data rate. Payload (Bytes) Retransmission limit is assumed infinity; each packet is (b) MIMO:SM Average Delay (PL=32 Bytes) sent as much as needed to get to the destination. The standard 250 1Mbps (SNR=6dB) default value is 7 [2], but loosening this limitations helps to 1Mbps (SNR=20dB) enhance reliability of the network to maximum (100%). 2Mbps (SNR=6dB) 200 2Mbps (SNR=20dB) Reliability is defined as the percentage of packets reaching to the destination. 150 Packet data payload for monitoring and control applications is typically short; 8, 32, 128 and 256 bytes are Delay (mS) 100 simulated and compared. Quasi-static flat fading multipath Rayleigh channel model 50 is used; channel condition doesn’t change during each packet transmission time, and changes for the next. 0 Saturated traffic mode is considered: all 4 access 5 10 15 20 25 Number of Nodes (n) categories in all nodes always have a packet waiting to be (c) sent. It is a good indicator of network performance under Fig. 1. MIMO Spatial multiplexing (MIMO:SM) average EDCA AC0 heavy traffic load and shows the upper bound to delay, saturation delay for 1/2Mbps data rate and in (a) different SNR’s, (b) Packet Error Rate (PER) and Bit Error Rate (BER) [4], [5]. diffrenent packet payload, (c) different number of competting nodes Both causes of packet loss are considered: collision and channel error. Each of them initiates the backoff process in B. EDCA Delay IEEE 802.11e CSMA/CA [1]. Fig. 1 compares the packet delay for different Transceivers have 2 Receive (Rx) and Transmit (Tx) configurations. It can be seen in fig. 1a that at high Signal to antennas (2×2 MIMO). Maximum Likelihood (ML) spatial Noise Ratios (SNR) (above 12dB), 2Mbps data rate has multiplexing (SM) receiver, with 2 spatial paths SM order of lower packet delay, but at medium (6-12dB) and low (below 2) has been simulated, which results in Spatial Diversity (SD) 6dB) SNR’s delay is worse than 1Mbps data rate, and gain of 2, improving the error rate [10]. increases exponentially. At low SNR’s, the packet payload

319 International Journal of Information and Electronics Engineering, Vol. 2, No. 3, May 2012 impacts the delay and it is increased exponentially. Fig.1b generally bounded. Fig. 3a shows that at medium and low studies the effect of packet payload. It is seen that at high SNR’s, retransmission count increases exponentially with SNR’s, as expected, high data rate outperforms the low one. increasing packet payload and data rate. Fig. 3b and c show But at medium and low SNR’s, 1Mbps delay increases that at high SNR’s or low data rate, packet payload and linearly with packet payload, whereas 2Mbps delay increases number of competing nodes have negligible effect on exponentially. It can be noticed in fig. 1c that the slope of retransmission count. 2Mbps data-rate delay at low SNR’s delay increase with respect to node density (n) at low SNR’s increases exponentially with payload and/or node density. is much higher for 2Mbps data rate. Fig. 2 compares different Fig. 4 compares the maximum retransmission counts for configurations from maximum packet delay perspective. As different configurations and has similar results as fig. 5. It can be seen, the slope of increase is higher with respect to Fig. can be inferred from theses simulations that, fixing 1. retransmission limit to its default value (7) has destructive effects on system reliability in sensor and actuator C. EDCA Retransmission Count applications due to low SNR environment. Hence, it should MIMO:SM Maximum Delay(n=15) 1400 be disabled, or for better performance, dynamically adapted 1Mbps (PL=8 Bytes) 1Mbps (PL=32 Bytes) to the network load, SNR and node density. Fig. 5 and 6 1200 2Mbps (PL=8 Bytes) 2Mbps (PL=32 Bytes) compare retransmission probability density for 8-byte and 1000 32-byte payload respectively. Fig.7 compares BER and PER for different configurations. The higher the data-rate and 800 packet payload, the higher is the error rate.

600 MIMO:SM Average Retransmission Count (n=15) Delay (mS) 70 1Mbps (PL=8 Bytes) 400 1Mbps (PL=32 Bytes) 60 1Mbps (PL=128 Bytes) 1Mbps (PL=256 Bytes) 2Mbps (PL=8 Bytes) 200 50 2Mbps (PL=32 Bytes) 2Mbps (PL=128 Bytes) 0 40 2Mbps (PL=256 Bytes) 2 4 6 8 10 12 14 16 18 20 SNR (dB) 30 (a) Retransmission Count MIMO:SM Maximum Delay (n=15) 20 5000 1Mbps (SNR=6dB) 4500 1Mbps (SNR=20dB) 10 2Mbps (SNR=6dB) 4000 2Mbps (SNR=20dB) 0 2 4 6 8 10 12 14 16 18 20 3500 SNR (dB) 3000 (a)

2500 MIMO:SM Average Retransmission Count (n=15) 35

Delay (mS) 2000 1Mbps (SNR=6dB) 1Mbps (SNR=20dB) 30 1500 2Mbps (SNR=6dB) 2Mbps (SNR=20dB) 1000 25 500

20 0 0 50 100 150 200 250 300 Payload (Bytes) 15 (b) Retransmission Count MIMO:SM Maximum Delay (PL=32 Bytes) 10 1400 1Mbps (SNR=6dB) 5 1Mbps (SNR=20dB) 1200 2Mbps (SNR=6dB) 0 2Mbps (SNR=20dB) 0 50 100 150 200 250 300 1000 Payload (Bytes) (b)

800 MIMO:SM Average Retransmission Count (PL=32 Bytes) 14

600 Delay (mS) 12

1Mbps (SNR=6dB) 400 10 1Mbps (SNR=20dB) 2Mbps (SNR=6dB) 2Mbps (SNR=20dB) 200 8

6 0 5 10 15 20 25 Number of Nodes (n) Retransmission Count 4 (c) 2 Fig. 2. MIMO Spatial multiplexing maximum EDCA AC0 saturation delay

for 1/2Mbps data rate and in (a) different SNR’s, (b) diffrenent packet 0 5 10 15 20 25 payload, (c) different number of competting nodes Number of Nodes (n)

(c) Fig. 3 compares the average number of retransmissions Fig. 3. MIMO Spatial multiplexing average number of AC0 retransmissions needed to have 100% reliability in different configurations. for 1/2Mbps data rate and in (a) different SNR’s, (b) diffrenent packet As expected, retransmission counts at high SNR’s are payload, (c) different number of competting nodes

320 International Journal of Information and Electronics Engineering, Vol. 2, No. 3, May 2012

MIMO:SM Maximum Retransmission Count (n=15) Retransmissions Probability Density (2Mbps, SNR=6dB, n=15, PL=8, Bytes) 300 0.1 1Mbps (PL=8 Bytes) 1Mbps (PL=32 Bytes) 0.09 250 1Mbps (PL=128 Bytes) 1Mbps (PL=256 Bytes) 0.08 2Mbps (PL=8 Bytes) 0.07 200 2Mbps (PL=32 Bytes) 2Mbps (PL=128 Bytes) 2Mbps (PL=256 Bytes) 0.06

150 0.05

0.04

100 Probability Density Retransmission Count 0.03

50 0.02

0.01

0 2 4 6 8 10 12 14 16 18 20 0 -10 0 10 20 30 40 50 60 SNR (dB) Number of Retransmissions (a) (b) MIMO:SM Maximum Retransmission Count (n=15) Fig. 5. MIMO Spatial multiplexing retransmission-count probability density 200 1Mbps (SNR=6dB) for 8-byte packet payload (a) 1mbps data rate (b) 2mbps data rate 180 1Mbps (SNR=20dB) 2Mbps (SNR=6dB) Retransmissions Probability Density (1Mbps, SNR=6dB, n=15, PL=32, Bytes) 160 2Mbps (SNR=20dB) 0.4

140 0.35 120 0.3 100

80 0.25

Retransmission Count 60 0.2

40 0.15 Probability Density 20 0.1 0 0 50 100 150 200 250 300 Payload (Bytes) 0.05 (b) 0 MIMO:SM Maximum Retransmission Count (PL=32 Bytes) 0 1 2 3 4 5 6 7 8 9 10 11 80 Number of Retransmissions (a) 70 1Mbps (SNR=6dB) Retransmissions Probability Density (2Mbps, SNR=6dB, n=15, PL=32, Bytes) 1Mbps (SNR=20dB) 0.09 60 2Mbps (SNR=6dB) 2Mbps (SNR=20dB) 0.08 50 0.07

40 0.06

30 0.05 Retransmission Count

20 0.04

10 Probability Density 0.03

0.02 0 5 10 15 20 25 Number of Nodes (n) 0.01

(c) 0 -10 0 10 20 30 40 50 60 70 80 90 Fig. 4. MIMO Spatial multiplexing maximum number of AC0 Number of Retransmissions retransmissions for 1/2mbps data rate and in (a) different SNR’s, (b) (b) diffrenent packet payload, (c) different number of competting nodes Fig. 6. MIMO Spatial multiplexing retransmission-count probability density for 32-byte packet payload (a) 1Mbps data rate (b) 2mbps data Retransmissions Probability Density (1Mbps, SNR=6dB, n=15, PL=8, Bytes) rate 0.45

MIMO:SM Bit Error Rate (n=15) 0.4 0.08 1Mbps (PL=8 Bytes) 0.35 0.07 1Mbps (PL=32 Bytes) 1Mbps (PL=128 Bytes) 1Mbps (PL=256 Bytes) 0.3 0.06 2Mbps (PL=8 Bytes) 2Mbps (PL=32 Bytes) 0.25 0.05 2Mbps (PL=128 Bytes) 2Mbps (PL=256 Bytes)

0.2 0.04 Bit Rate Error Probability Density Probability 0.15 0.03

0.1 0.02

0.05 0.01

0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 2 4 6 8 10 12 14 16 18 20 SNR (dB) Number of Retransmissions (a) (a)

321 International Journal of Information and Electronics Engineering, Vol. 2, No. 3, May 2012

MIMO:SM Packet Error Rate (n=15) [7] R. Moraes, P. Portugal, and F. Vasques, “Simulation analysis of the 1 1Mbps (PL=8 Bytes) IEEE 802.11e EDCA protocol for an industrially-relevant real-time 0.9 1Mbps (PL=32 Bytes) communication scenario,” in Proc. 11th IEEE Int. Conf. on Emerging 1Mbps (PL=128 Bytes) 0.8 1Mbps (PL=256 Bytes) Technologies and Factory Automation (ETFA 2006), Czech Republic, 2Mbps (PL=8 Bytes) 2006, pp. 202-209. 0.7 2Mbps (PL=32 Bytes) 2Mbps (PL=128 Bytes) [8] N. Chendeb Taher, Y. Ghamri Doudane, B. El Hassan, “A complete 0.6 2Mbps (PL=256 Bytes) and accurate analytical model for 802.11e EDCA under saturation 0.5 conditions,” in Proc. IEEE/ACS Int. Conf. on Computer Systems and 0.4 Applications (AICCSA 2009), Rabat, 2009, pp. 800-807. Packet Error Rate Error Packet 0.3 [9] G. Cena, L. Seno, A. Valenzano, and C. Zunino, “Evaluation of

0.2 real-time communication performance in QoS-enabled infrastructure WLANs,” in Proc. 14th IEEE Int. Conf. on Emerging Technologies and 0.1 Factory Automation (ETFA 2009), Mallorca, 2009, pp. 1-7. 0 2 4 6 8 10 12 14 16 18 20 [10] Tolga M. Duman and Ali Ghrayeb, Coding for MIMO Communication SNR (dB) Systems, Chichester, U.K.:John Wiley & Sons Ltd, 2007, ch. 5. (b) Fig. 7. MIMO Spatial multiplexing error performance for 1/2mbps data rate (a) bit error rate (BER) (b) packet error rate (PER) Mohsen Maadani received two B.S. degrees in biomedical engineering and electronic engineering form Amirkabir University of Technology (AUT), Tehran, Iran and M.S. degree in electronic engineering from Sharif University of Technology IV. CONCLUSION AND FUTURE WORK (SUT), Tehran, Iran in 2003 and 2006 respectively. He was a Senior Technical Engineer in In this paper, saturation delay and reliability of EDCA communication and R&D department of Iran Control mechanism for SM-enabled IEEE 802.11 based wireless and Communication Systems Supply (ICS) sensor and actuator networks has been analyzed. Simulation Company, Tehran, Iran from Jan 2006 to Sep 2007 and has been a faculty member at Islamic Azad University (IAU), Shahr-e-Qods Branch and also a results show that simple modulation schemes with spectral PHD candidate in electronic engineering at Amirkabir University of efficiency of 1 usually outperform high-rate ones, due to the Technology (AUT) since 2007. His recent research activities are mainly noisy and short packet payload nature of industrial networks. focused on the Physical and Medium Access Control (MAC) layers of Also for reliable applications, retransmission limit must be wireless networks, including Wireless Industrial Networks (WIN), Wireless Sensor Networks (WSN), Wireless Local Area Networks canceled, loosened or dynamically adopted to the network (WLAN), Wireless Personal Area Networks (WPAN), Multiple Input condition. Otherwise, packet loss ratio will increase beyond Multiple Output (MIMO) Communication Systems, Cross-Layer Design, acceptable limit. Our future work is to assess using more and adaptive Quality of Service (QoS) provisioning schemes. Mr. Maadani is a member of IEEE and IACSIT. sophisticated spatial multiplexing codes, and adaptive schemes. Seyed Ahmad Motamedi received the B.S. degree in electronic engineering from Amirkabir University of Technology (AUT), Tehran, Iran, in 1979. He EFERENCES R received the M.S. degree in computer hardware in [1] IEEE Standard for Information technology- Telecommunications and 1981 and PHD degree in information systems information exchange between systems- Local and metropolitan area (computer hardware) in 1984, both from University networks- Specific requirements, Part 11: Wireless LAN Medium of Pierre & Marie Curie (Paris VI), France. Access Control (MAC) and Physical Layer (PHY) Specifications, Currently he is a Full Professor in Electrical IEEE 802.11 Standard-2007. Engineering Department of Amirkabir University of [2] IEEE Standard for Information Technology- Telecommunications Technology (AUT) and has been a faculty member of this institute since and information exchange between system- Local and metropolitan 1984. His research interests include, Parallel Processing, Image Processing, area network--Specific requirements Part 11: Wireless LAN Medium Microprocessor Systems, Wireless Industrial Networks, Wireless Sensor Access Control (MAC) and Physical Layer (PHY) specifications Networks, Automation and Biomedical Engineering. Amendment 5: Enhancements for Higher Throughput, IEEE 802.11n Prof. Motamedi was the president of Iranian Research Organization for -2009. Science and Technology (IROST) from 1986 to 2001. He is the head of [3] A. Willig, “Recent and Emerging Topics in Wireless Industrial High-Speed Processing Research Center, AUT, Tehran, Iran. He has been Communications: A Selection,” IEEE Trans. On Industrial Informatics, the author of several books and published a lot of scientific papers in Vol. 4, No. 2, pp. 102-124, May 2008. international conferences and journals [4] M. Maadani, S. A. Motamedi, and M. Mohammadi Noshari, "Delay Analysis and Improvement of IEEE 802.11e-Based Soft-Real-Time Wireless Industrial Networks using an Open-Loop Spatial Mohammadreza Soltani received his B.S. degree Multiplexing Scheme," in Proc. IEEE International Symposium on in from Guilan University, Rasht, Iran and M.S. Computer Networks and Distributed Systems (CNDS 2011), Tehran, degree from Amirkabir University of Technology 2011, pp. 17 - 22. (AUT), Tehran, Iran in 2009 and 2011 respectively [5] G. Cena, L. Seno, A. Valenzano, and C. Zunino, “On the performance both in electrical engineering. of IEEE 802.11e wireless infrastructures for soft-real-time industrial His research interests are in the areas of Wireless Ad applications,” IEEE Trans. on Industrial Informatics, Vol. 6, No. 3, pp. Hoc and Sensor Networks with emphases on 425-437, Aug 2010. reliability and energy-efficient algorithms in routing [6] G. Cena, I. Cibrario Bertolotti, A. Valenzano, and C. Zunino, protocols, cross layer design protocol and “Evaluation of response times in industrial WLANs,” IEEE Trans. on application of game theory in wireless Networks. Industrial Informatics, Vol. 3, No. 3, Aug 2007, pp. 191-201.

322