Automating Wireless Sensor Network Deployment and Replacement in Pipeline Monitoring Ted Tsung-Te Lai Albert Wei-Ju Chen Kuei-Han Li Polly Huang Hao-Hua Chu National Taiwan University Ted Tsung-Te Lai Albert Wei-Ju Chen Kuei-Han Li Department of Computer Science and Information Engineering Polly Huang Graduate Institute of Networking and Multimedia Department of Electrical Engineering Hao-Hua Chu Graduate Institute of Networking and Multimedia Department of Computer Science and Information Engineering Outline Motivation TriopusNet System Design Evaluation Limitations Related Work Conclusion Water pipelines are everywhere people live Pipelines carry important resources (gas, oil…etc.) Pipeline monitoring is essential-clean water •Motivation leaking leaking Water contamination (Boston, 2010) Difficult sensor deployment-traditional monitoring WSN challenges (Deployment and maintenance) • Deployment challenges – Difficult to access pipelines to place sensors (often hidden inside walls or underground) – May need to break pipes to install sensors inside • Maintenance challenge – Difficult to replace out-of-battery sensors • Real pipeline environment – Difficult to ensure network connectivity during sensor placement and replacement Research question • Can we automate WSN sensor placement and replacement in pipeline? – While minimize the number of sensor nodes – Good sensing and networking coverage • Reduce the human effort bottleneck for long-term, large-scale WSN deployment & maintenance. The system involves the following: 1. Preparation Step • Knowing the spatial topology(turning faucets on one after another). 2. Sensor Deployment Step • Compute deployment location then send “release” message and position to node. 3. Sensor Latching Step • Compute location, attach itself, completion message. 4. Sensor Replacement Step • Consume battery power during the data collection phase. • Detach itself, go to faucet, exit. Single-Release Point the enabling concept Place sensors at a single release point Sensors automatically place themselves in the pipes Single-release point How to realize single-release point? • Sensor placement – – – – Mobile sensors Sensor latch mechanism Sensor placement algorithm Sensor localization • Sensor replacement – Sensor replacement algorithm Limitations 1. The spatial topology of pipeline must be known. 2. Manual effort is required to open faucets.(at the beginning, at battery replacement) 3. Current sensor measures 6 cm in diameter. Outline Motivation TriopusNet System Design Evaluation Limitations Related Work Conclusion TriopusNet automate WSN deployment in pipeline Triopus node three arms for latching Single-release point Gateway node Gateway node Gateway node TriopusNet automate WSN deployment in pipeline • Sensor placement – – – – Mobile sensors Sensor latch mechanism Sensor placement algorithm Sensor localization • Sensor replacement – Sensor replacement algorithm Mobile sensor (components) Sensor mote(Kmote) Actuator(motor) pull/push a mechanical arm Localization sensors water pressure + Vertical Pipe gyro horizontal pipe Water proof case Mobile sensor (kmote) + = Kmote CPU board (data processing) USB board (program uploading) • A TelosB-like platform, TinyOS compatible • Smaller form-factor, only CPU board is needed Mobile sensor (latch & delatch mechanism) Linear actuator, off-the-shelf from market A motor with gear inside to control the arm Spec: • Stroke: 2cm • Weight: 15gram • Arm extending speed: 2cm/sec 2cm 1cm 0cm Prototype #1 (8cm diameter) Prototype #2 • One motor driving the three arms. • Replace 3 AAA with lithium battery. Prototype #2 (6cm diameter) Sensor placement algorithm • Where are the optimal locations to place sensors in pipes (after releasing them from the single-release point)? – Networking coverage • Interconnectivity among all nodes – Sensing coverage • Each pipe segment has at least one sensor – Minimize # of sensor nodes for deployment Sensor placement algorithm root water inlet branch 2 n7 branch 1 faucet 1 branch 3 n6 n1 faucet 4 faucet 3 n4 n5 faucet 2 n2 n3 Sensor placement algorithm root water inlet branch 2 n7 branch 1 faucet 1 branch 3 n6 n1 faucet 4 faucet 3 n4 n5 faucet 2 n2 n3 Sensor placement algorithm root water inlet branch 2 n7 branch 1 faucet 1 branch 3 n6 n1 faucet 4 faucet 3 n4 n5 faucet 2 n2 n3 Sensor placement algorithm root water inlet branch 2 n7 branch 1 faucet 1 branch 3 n6 n1 faucet 4 faucet 3 n4 n5 faucet 2 n2 n3 Sensor placement algorithm root Post-order traversal : n1 -> n2 -> … n7 n7 n6 n1 n4 n2 n5 n3 Sensor placement algorithm root Post-order traversal : n1 -> n2 -> … n7 n7 1st n6 n1 n4 n2 n5 n3 Sensor placement algorithm root Post-order traversal : n1 -> n2 -> … n7 n7 1st n6 n1 2nd n2 n4 n5 n3 Sensor placement algorithm root Post-order traversal : n1 -> n2 -> … n7 n7 1st n6 n1 2nd n2 n4 3rd n3 n5 Sensor placement algorithm root Post-order traversal : n1 -> n2 -> … n7 n7 1st n1 4th 2nd n2 n4 n6 3rd n3 n5 Sensor placement algorithm root Post-order traversal : n1 -> n2 -> … n7 n7 1st n1 4th 2nd n2 n4 n6 3rd n3 5th n5 Sensor placement algorithm root Post-order traversal : n1 -> n2 -> … n7 n7 6th 1st n1 4th 2nd n2 n4 n6 3rd n3 5th n5 Sensor placement algorithm root Post-order traversal : n1 -> n2 -> … n7 7th n7 6th 1st n1 4th 2nd n2 n4 n6 3rd n3 5th n5 Sensor placement algorithm Post-order traversal : n1 -> n2 -> … n7 root Reasons: 1. Assure nodes cover all pipes 2. Allow blockage-free movement (bottom-up placement) 7th n7 6th 1st n1 4th 2nd n2 n4 n6 3rd n3 5th n5 Sensor placement algorithm Single-release point Gateway node Testing packet received ratio Bad link quality Good link quality, placement completed Gateway node Gateway node Sensor localization Pressure graph • Previous PipeProbe system – cm-level positional accuracy • Vertical pipe location – Water pressure changes at different height levels • Horizontal pipe location – Node distance = node velocity * node flow time • Pipe turn detection – Gyroscope Data Collection • Collection Tree Protocol (CTP) in TinyOS • Multi-sink tree to balance network load (reduce the hope count and packet loss) Single-release point Gateway node Gateway node Gateway node Sensor replacement algorithm Single-release point Gateway node Gateway node Low Battery… Gateway node Outline Motivation TriopusNet System Design Evaluation Limitations Related Work Conclusion Testbed Testbed spatial layout Single-release point 150cm 200cm 200cm 200cm 200cm 200cm Evaluation metrics • Automated sensor placement – # Nodes for pipeline deployment – Data collection rate – Energy consumption • Automated sensor replacement – Data collection rate Experimental procedure (4 test scenarios) Single-release point 5 tests for each scenario gateway Scenario 3 Scenario 2 Scenario 4 Scenario 1 gateway gateway # Deployed Nodes (Static v.s. TriopusNet deployment) Real node location of three test runs from scenario 4. It shows the dynamic of each deployment. Static (90cm) TriopusNetA TriopusNetB TriopusNetC # Automated Sensor Deployment Avg # of nodes deployed -Static: 7.5 -TriopusNet: 4.4 Avg. node-to-node distance: 173cm Std: 58cm • The overall large variation implies that the Radio range varies significantly from location to location. Avg. node-to-node distance Avg. node-to-node distance Avg. node-to-node distance Avg. node-to-node distance • I-shape radio signal travel through water which absorb energy and limits its range. Data collection rate-cumulative density function Each node sent 1000 packets to gateway -80% nodes achieve 99% packet receive rate -All nodes > 86.5% rate CDF of Positional Errors • Overall median error 7.14 cm • 90% of errors are less than 20.45 cm Location Estimates: 18 20 20 30 • Node positional accuracy is important for achieving sensing coverage in node deployment. Energy consumption (node placement) The energy consumed by a single act of latching is 1.01 W ,2 seconds The average of latching is 2.35 90% required less than 5 Evaluation metrics • Automated sensor placement – # nodes for sensing/networking coverage – Data collection rate – Energy consumption • Automated sensor replacement – Data collection rate Test scenario and result for replacement Data collection rate Initial deployment After replacement Without replacement 0.989 0.984 0.81 The effectiveness of the automated replacement Set these two nodes to low battery level and trigger replacement The reason of high data loss rate: 1-some sensors change route 2-isolated nodes report zero Outline Motivation TriopusNet System Design Evaluation Limitations Related Work Conclusion Limitation: Lack automatic faucets automatic faucet The TriopusNet gateway control each faucet by sending signals to the sensor trigger node. Limitation: Node size Limitation: Node size Single-release point Low Battery… Outline Motivation TriopusNet System Design Evaluation Limitations Related Work Conclusion Mobile WSN Deployment Breadcrumb System Liu SensorFly System Purohit PipeNet (pipeline monitoring) Detect and localize leakage by pressure and ultrasonic sensors NAWMS (water flow sensing) HydroSense (Ubicomp’09, water event sensing) Single-point pressure-based sensor of water usage toilet kitchen sink shower PipeProbe (determining the spatial topology) Outline Motivation TriopusNet System Design Evaluation Limitations Related Work Conclusion Conclusion TriopusNet: automating WSN deployment and replacement in pipeline monitoring Automated sensor placement and replacement to reduce human deployment and maintenance effort: mobile sensors with self-latching mechanism from a single-release point Results show smaller number of sensor nodes with good sensing/networking coverage Thank You