An Experimental Evaluation of Selective Cooperative Relaying for Industrial Wireless Sensor Networks Nikolaj Marchenko, Torsten Andre, G¨ unter Brandner, Wasif Masood, and Christian Bettstetter Institute of Networked and Embedded Systems University of Klagenfurt, Austria
Quick Intro R S D 1/21
Quick Intro R S D 2/21
Quick Intro ◮ Lots of theoretical work on cooperative relaying ◮ Some experimental studies (40+ articles): 1 Software-Defined Radios: Mostly PHY Layer, few nodes 2 Wireless Sensor Nodes: MAC/Network layer, many nodes 3/21
Application in Industrial WSN ◮ Monitoring and control of production processes ◮ Harsh environment for wireless signal propagation ◮ Very strict requirements on link reliability and delay ◮ Standards: WirelessHART, ISA100.11a, Zigbee IA Profile 4/21
Relay Selection Relay Selection 1 Which metrics to use for relay selection? ◮ Channel quality info, residual battery life, etc. 2 How selection is coordindated? ◮ Signaling messages, contention mechanism, etc. 3 How often a relay is updated? ◮ Update requirements and policy. R S D 5/21
Relay Selection I. Periodic Relay Selection A relay is selected strictly at periodic time intervals T sel ◮ random contention of candidates in window w . ◮ based on current Link Quality Indicators S − R and R − D . selection retransmission D S R i S R i D err DATA ( mc) S_RREQ (bc) T ACK DATA rand(0, w ) ACK (mc) R_CAND ACK DATA (mc) D_RSEL T ACK ACK (mc) R_RSEL ACK 6/21
Relay Selection II. Adaptive Selection ◮ A new relay selection when in a window W a more than ε a ACK s are lost. ◮ Selection and retransmission procedure same as periodic 7/21
Relay Selection III. Reactive Selection Selection is performed after each missing ACK for direct S - D transmissions among nodes that 1 have received the packet correctly 2 have a good channel to the destination DATA delivery by S fails DATA is delivered by S S R i D S R i D DATA ( bc) err DATA ( bc) no err T ACK ACK (bc) T ACK S_RREQ ( bc) relay selection DATA (bc) rand(0, w ) R_CAND ACK (bc) err D_RSEL ACK err T ACK DATA S_RREQ (bc) ACK ( mc) rand(0, w ) ACK ACK 8/21
Test Environment 9/21
Test Environment 10/21
Hardware ◮ Crossbow TelosB ◮ TinyOS implementation ◮ Transmission: 2.4 GHz, 256 kbit/s, TxPower: -4 dBm 11/21
Layout ◮ 7 nodes, 6 tested links 25m D 60m ◮ Two kind of experiments: 1 Trace-based analysis on individual link: Node 6 transmits to D every 160 ms, other nodes retransmit 2 Explicit experimental comparison over all links 12/21
Performance Results: Trace-Based Analysis 1 Periodic selection: every 200 packets 2 Adaptive selection: if error rate > 10% for last 50 packets. Number of neighboring nodes 1 node 2 nodes 3 nodes 4 5 1 0.4 0.9 delivery ratio 0.8 reactive 0.7 periodic adaptive 0.6 time diversity 0.5 direct transmissions only 0.4 0 5 10 15 20 25 30 node combination id 13/21
Performance Results: Trace-Based Analysis Periodic selection number of selection per 100 data pkts 0.95 10 2 0.9 # relays 10 1 1 delivery ratio 0.85 2 3 1 0.8 4 10 0 2 5 3 0.75 5 10 − 1 0.7 10 0 10 1 10 2 10 0 10 1 10 2 selection period in pkts selection period, data packets 14/21
Performance Results: Trace-Based Analysis Adaptive selection: Window Adaptive selection: Error rate 0.9 0.9 0.85 total delivery ratio N = 3 , 4 , 5 N = 2 0.85 delivery ratio 1 0.8 2 0.8 3 0.75 5 0.75 N = 1 0.7 0 0.2 0.4 0.6 0.8 1 error rate threshold, ε a 0.7 0 50 100 150 200 250 300 10 2 window size, W a selections per 100 data packets 10 2 10 1 selections per 100 data packets 3 1 4 2 10 0 5 10 1 10 − 1 1 2 10 − 2 3 10 0 5 10 − 3 0 0.2 0.4 0.6 0.8 1 10 − 1 error rate threshold, ε a 0 50 100 150 200 250 300 window size, W a 15/21
Direct Comparison 160ms 160ms 160ms 160ms time div. periodic adaptive reactive time div. time Exp.2. Test:3081 1 0.9 0.8 Delivery ratio per sample 0.7 0.8 0.6 0.5 direct 0.4 periodic adaptive 0.3 reactive 0.2 0.1 0 0 500 1000 1500 2000 Sample # 16/21
Performance Results: Direct Comparison ◮ Measurements on 3 days, each 12 hours. ◮ Total 810K DATA packets sent by source nodes ◮ 33K on each link and each scheme, over 6 hours time. ◮ Periodic selection: T sel = 400 · 160 ms = 64 sec ◮ Adaptive selection: W a = 100, ε a = 0 . 1 Table: Mean Results over the Network direct time div. periodic adaptive reactive delivery ratio, % 81.2 85.7 96.9 97.9 98.9 selections per 100 pkts - - 1.08 1.11 22.7 number of candidates - - 3.69 3.86 3.43 selection success, % - - 94 91 92 relaying success,% - - 78 82 95 17/21
Performance Results: Direct Comparison Delivery ratio in a sample Selections in a sample 10 0 1 0.9 0.8 total delivery ratio 0.7 10 − 1 0.6 single direct transmission cdf 0.5 time diversity time diversity 0.4 periodic periodic adaptive 10 − 2 0.3 adaptive reactive reactive 0.2 0.1 0 10 − 3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 delivery ratio by a single direct transmission in a sample delivery ratio in a sample 18/21
Performance Results: Direct Comparison Delivery ratio in a sample 1 0.95 0.9 0.9 0.85 0.8 time diversity cdf 0.75 periodic 0.7 adaptive reacitve 0.65 0.6 0.55 0.5 10 0 10 1 10 2 number of retransmission rounds 19/21
Conclusions 1 Cooperative relaying provides up to 99% delivery ratio 2 Short-term outages are also avoided 3 Adaptive selection provides best tradeoff between delivery ratio and selections overhead Publications ◮ N. Marchenko, et al. An Experimental Study of Selective Cooperative Relaying in Industrial Wireless Networks. Under review in IEEE Trans. Industrial Informatics , 2013. ◮ T. Andre, et al. WiNMee Workshop, May 2013. ◮ T. Andre, et al. GLOBECOM’12 . 20/21
Finally the Final Slide Thank You for Your Attention! 21/21
Recommend
More recommend
