A Distributed Sensor Relocation Scheme for Environmental Control

1. A Distributed Sensor Relocation Scheme for Environmental Control Michele Garetto, Università di Torino Marco Gribaudo, Università di Torino Carla-Fabiana Chiasserini, Politecnico di Torino Emilio Leonardi, Politecnico di Torino

2. Outline • Introduction to the problem • Our solution • Performance evaluation • Conclusions

3. Mobile sensor networks ? • Traditionally, sensor networks have been assumed to be static… • …but mobile sensor networks are becoming real • …with many promising applications

4. Network scenario • Large number of self organizing, unattended mobile sensors with actuators (micro-robots) • Limited memory/computing capability • Short radio range • Energy-limited (battery operated) • No GPS

5. Deployment and Relocation problem • How to achieve coordinated motion of the nodes to improve area coverage and/or relocate upon occurrence of events? ?

6. Our objective • Design a unified algorithm to jointly achieve network deployment and relocation • Fully distributed solution: no centralized control, no coordination/communication between distant nodes • Meet the constraints of the nodes: limited energy, computation, communication capabilities • No need of absolute node localization (only relative position of neighboring nodes)

7. Our approach • Consider large-scale relocation of the nodes, no fine-grained details (e.g.: filling holes) • Take a macroscopic view on how network behaves as a whole • Each nodes acts an independent agent and interacts with neighbors according to a simple set of rules • Exploit swarm intelligence to achieve self-deployment and relocation as emergent behavior

8. Our proposed solution • Customized virtual forces approach • The virtual force acting on bode i at time t is: Friction forces (needed to stabilize the network) static +viscous Resultant of attractive/repulsive forces exchanged with neighboring nodes j Potential force activated only when an event is sensed by the node

9. Attractive/repulsive forces • Needed to achieve target distance (Dm) between nodes while maintaining network connectivity (no boundaries) • We need to estimate distance (from RSSI) and direction of arrival (DoA) of signals received by each neighbor errors considered: distance (±5%), angle (±10°)

10. Selection of active neighbors 60°- Δ° Communication range

11. Rs Self-deployment • Starting from any (connected) initial topology, the equilibrium configuration tends to a regular triangular lattice … … Dm Optimal coverage when … …

12. Example of self-deployment n = 400 nodes

13. Our scheme – no errors Our scheme – with error Self deployment: coverage results Rs = 1 n = 400 100 Perfect triangular lattice Random placement 95 90 85 Coverage Percentage 80 75 70 65 2.4 1.2 1.4 1.6 1.8 2 2.2 Dm

14. Initial topology Final topology Performance evaluation • Metrics: • Time taken to reach final configuration • Total movement of the nodes (to save energy) • We compare our scheme with the optimum centralized solution reaching the same final configuration: • Nodes move at the maximum speed all the time • The selection of which node goes where is done solving a minimum Weight Matching (mWM) problem

15. 300 250 200 150 100 algorithm - G = 0.01 algorithm - G = 0.001 50 mWM mWM 0 0 400 800 1200 1600 2000 Time Comparison with optimum centralized solution (mWM) 350 300 250 Total Movement 200 150 100 50 0 0 100 200 300 400 Time

16. Relocation upon occurrence of event • Nodes sensing an event are subject to an additional, constant force directed towards the event • The objective is to achieve a given node density around the event, possibly keeping a safe distance from it • Local density is obtained by dynamically tuning the intensity of the exchange forces among neighboring nodes

17. Example of event-based relocation

18. Performance evaluation • We compare again our distributed scheme with the optimal centralized one (mWM) which minimizes total node movement • We count how many nodes arrive at a given distance d from the event epicenter as a function of time

19. Comparison between our algorithm and mWM algorithm d < 18 400 mWM 350 300 250 d < 12 Number of Sensors 200 150 d < 9 100 50 0 0 500 1000 1500 2000 2500 3000 3500 4000 Time

20. Optimum relocation (mWM)

21. Limited event detection

22. Multiple concurrent events

23. Conclusions • We have proposed a distributed, unified solution for self-deployment and event-based relocation in mobile sensor networks • Simple local rules allow the network to behave as an intelligent swarm • Performance comparable with that achieved by centralized optimum solution

24. 400 R = 80 R = 40 350 R = 30 300 250 Number of Sensors 200 150 100 50 0 0 1000 2000 3000 4000 5000 Time

27. Results: coverage after deployment