1. Network and Systems Laboratory nslab.ee.ntu.edu.tw Branislav Kusy, Christian Richter, Wen Hu, Mikhail Afanasyev, Raja Jurdak, Michael Brunig, David Abbott, Cong Huynh, and Diethelm Ostry CSIRO ICT Centre, Australia Archiang 2011/06/13

2. Network and Systems Laboratory nslab.ee.ntu.edu.tw Outline Introduction Theoretical Basis of Diversity Hardware Platform Implementation Evaluation Conclusions

3. Network and Systems Laboratory nslab.ee.ntu.edu.tw Outline Introduction Theoretical Basis of Diversity Hardware Platform Implementation Evaluation Conclusions

4. Network and Systems Laboratory nslab.ee.ntu.edu.tw Introduction Low power wireless mesh is the most common communication architecture for WSNs today. State-of–the-art data collection protocols react quickly to changes in network connectivity and repair their routing state. Network congestion and increased packet collisions Link failures become more expensive to repair with increasing network size as the routing state may need to be propagated network-wide.

5. Network and Systems Laboratory nslab.ee.ntu.edu.tw Issues This paper improves primary radio link reliability to construct large reliable and efficient WSNs spanning regions. Selection of node locations to achieve a robust wireless mesh is a challenging problem. Cost increased by requiring additional hardware or time- consuming iterative deployment adjustment procedures.

6. Network and Systems Laboratory nslab.ee.ntu.edu.tw Proposed Architecture A new network architecture is proposed based on this radio communication diversity. A new platform with two independent radios operating at well- separated frequencies and spatially separated antennas. 2 IEEE 802.15.4 radio chips, operating in the 900MHz and 2400MHz bands Disadvantage Lower energy efficiency (33% more energy) Loss of compatibility with existing hardware Additional components cost

7. Network and Systems Laboratory nslab.ee.ntu.edu.tw Outline Introduction Theoretical Basis of Diversity Hardware Platform Implementation Evaluation Conclusions

8. Network and Systems Laboratory nslab.ee.ntu.edu.tw Signal Loss Multi-path propagation Spatially Selective fading The far scattering model The near scattering model Frequency selective fading Bounds on spatial and frequency fading Obstruction and environmental fading The line of sight propagation path between nodes is obstructed. Radius of Fresnel zone Interference 2400MHz:WiFi networks. 900MHz:Telemetry networks and cordless telephones.

9. Network and Systems Laboratory nslab.ee.ntu.edu.tw The far scattering model The near scattering model

10. Network and Systems Laboratory nslab.ee.ntu.edu.tw Frequency selective fading Frequency selective fading is caused by differences in the times of arrival of multiple radio signals traveling along different path. Bounds on spatial and frequency fading

11. Network and Systems Laboratory nslab.ee.ntu.edu.tw Diversity Design For multi-path In 802.15.4(2MHz BW) or WiFi(20MHz BW) An antenna separation: 1m Frequency separation: 100MHz For interference Frequency diversity is robust against interference if the sources of interference are uncorrelated at the multiple frequencies.

12. Network and Systems Laboratory nslab.ee.ntu.edu.tw Outline Introduction Theoretical Basis of Diversity Hardware Platform Implementation Evaluation Conclusions

13. Network and Systems Laboratory nslab.ee.ntu.edu.tw Hardware Platform Each board was configured to use a separate antenna and thus has its own RF connector. Antennas provide a 5dBi gain on 2400 MHz and 2.2dBi gain on 900 MHz.

14. Network and Systems Laboratory nslab.ee.ntu.edu.tw Outline Introduction Theoretical Basis of Diversity Hardware Platform Implementation Evaluation Conclusions

15. Network and Systems Laboratory nslab.ee.ntu.edu.tw Implementation Dual-band radio driver Separate the two radio stacks Execute the two drivers in parallel Link estimation layer 4 bit link estimator(4BLE) Data collection layer CTP-multi, the extension of CTP to multiple radio bands Low power listening Time of idle listening is double Energy consumption increases 3% - 33%

16. Network and Systems Laboratory nslab.ee.ntu.edu.tw Outline Introduction Theoretical Basis of Diversity Hardware Platform Implementation Evaluation Conclusions

17. Network and Systems Laboratory nslab.ee.ntu.edu.tw Characterization of Diversity Outdoor open space experiment Line-of-sight deployment, and no in-band interference 900m transmission range, 10 packets per second, and 250kbps data rate Improve PRR of these links by as much as 84% Indoor office space experiment Evaluate the benefits of frequency diversity Verify equal RSSI in both bands 100 locations 10 packets per second

18. Network and Systems Laboratory nslab.ee.ntu.edu.tw Channel diversity Channel diversity is insufficient in a number of scenarios Channel hopping provides limited benefits when multipath fading affects the radio signals and is far less useful in avoiding external interference.

19. Network and Systems Laboratory nslab.ee.ntu.edu.tw Data Collection Performance 30 nodes in 400m X 450m Max distance between neighbor nodes is 80m 90 bytes per packet, 4kbps 5 seconds inter-packet interval

20. Network and Systems Laboratory nslab.ee.ntu.edu.tw Comparison of dual to single band networks

21. Network and Systems Laboratory nslab.ee.ntu.edu.tw Evaluation of network diversity

22. Network and Systems Laboratory nslab.ee.ntu.edu.tw Energy Overhead of Multiple Bands Low power listening Current Sleep: 300uA Active: 14.1mA (RF212) 19.1mA (RF230) CTP uses unicast transmissions to deliver data packets to the base station and control traffic is a small fraction of the data traffic. Energy overhead

23. Network and Systems Laboratory nslab.ee.ntu.edu.tw Outline Introduction Theoretical Basis of Diversity Hardware Platform Implementation Evaluation Conclusions

24. Network and Systems Laboratory nslab.ee.ntu.edu.tw Conclusions This work improves reliability of primary radio links by radio diversity. Multiple antennas are mounted on each node to gain additional resilience against multipath fading. Using low power listening, energy overhead reduces to less than 33%. The future work includes improvement of the energy consumption of the diversity radio stack. Turning off the radios can decrease the overhead of idle listening.