Is There Light at the Ends of the Tunnel? Wireless Sensor Networks for Adaptive Lighting in Road Tunnels IPSN 2011 Sean Outline • • • • • • • • Goal Challenge Contribution System Architecture Hardware & Software Testbed Evaluation Conclusion Goal • WSN-based Close-loop adaptive lighting in road tunnel – Improve tunnel safety – Reduce power consumption • State-of-the-art solutions – Pre-set lighting based on date and time – Relying only on external sensor • Testbed evaluation • Real deployment – Project TRITon – 630m, two-way, two-lane tunnel Challenge • Peculiarities of Tunnels – – – – – – harsh environment, relatively studied on WSN o. Interference with WSN radio Vehicular traffic dirt and dust accumulation o. Occlusion & noise to light sensor Periodic tunnel cleaning direct sunlight Limited deployment & debugging Light variation • Need filtering – Better connectivity • Robustness • Packet collision Variation caused by vehicle Challenge • Real-world constraints – – – – – Extended lifetime : at least 1-year by tunnel operators WSN cannot fail due to continuous operation Sensed data must arrive timely Quality of sensing Integration with conventional, industrial-strength equipment Contribution • Verify WSN-based solution to adaptive lighting is feasible • Understand what extent the mainstream WSN technology can achieve • Real testbed implement • Gaining practical insight into tunnel scenario – Real-world lesson asset System Architecture • 3 components – An external sensor Measure the veil luminance Determine the legislated curve – A grid of light sensor along the tunnel length Compute error between legislated curve and actual lighting – A control algorithm Drive above error to zero HPS in Testbed LED for project Hardware & Software • Collection tree – Use LQI as path cost – Periodically reconstructed every 3min • Light Sensing – Average 4 sensor value into S(i) – Average all S(i) into S(all) – if |S(all) – S(i)| differs from S(all) by 50%, discard it – Recompute S(all) Testbed • • • • 40 nodes, 260m-long, two-way, two-lane tunnel PLC relies only on first 15 node 7-month experiments More dense than TRITon – 44 nodes, 630m • Light sensor sample every 5s, PLC collects data every 30s Evaluation • • • • • • Light adaptive effect Loss rate Timely delivery Resilience to gateway failures Retransmission cost Expected lifetime Light adaptive effect • Artificial step response Still follow the reference trend • Node position relative to lamps bears great influence • Behavior of other node is closer to node 7 than node 4 Light adaptive effect • Real-world reference Bound by the dynamic range of light actuator Only 150 lx maximum Loss rate Time spent transmitting and waiting for receiver to wake up becomes significant Timely delivery > 60s: PLC will loss more than one sample in its cycle 30~60s: PLC may loss a sample in its cycle Resilience to gateway failures Retransmission cost Expected lifetime • Battery discharge profile – Temperature – Voltage – Discharge current • Underestimate – Use average discharge current of 100mA – LPL-like MAC only consume a few mA • 250ms LPL is better – Power consumed in channel check – Packet strobe time Trade-off Conclusion • Reach the goal of close-loop adaptive lighting • Provide real-world insights and experience by using WSN in road tunnel Any Problem?