Optimal Self-placement of Heterogeneous Mobile Sensors in Sensor Networks Lidan Miao AICIP Research Oct. 19, 2004

2 Motivation Collaborative efforts of many sensors. The placement affect the effectiveness of sensor network: cost, coverage, network lifetime, etc. Self-deployable sensor network is required in hazardous environment or some military applications.

3 Literature review The sensor deployment has been an active research topic in sensor network. The environment is sufficiently known and under control. The deployment of mobile sensors based on centralized control. Self-deployable sensor networks. Assumption: Homogenous sensor platforms.

4 Objective Minimize energy consumption convergence time the number of sensor nodes Maximize network coverage

5 Proposed idea The sensor deployment contains two interrelated issues: Optimal placement design Convergence from initial state to optimal state by sensor movement The sensor locations are known before actual deployment.

6 Optimal placement algorithm Stimulated from the MSFA design. The sensing field is represented by a 2D array, each grid point denotes a sensor site. Different sensor platforms form a mosaic pattern.

7 Self-deployment algorithm Random movement Movement based on centralized control Swarm intelligence-based movement

8 Stopping criterion The optimal placement confines the sensor site of different sensor platforms. 3 sensor types 7 sensor types

9 Random Movement At each position, the sensor randomly choose one direction to go. It is very likely for one sensor to visit the same location many times. It is not energy efficient. Low convergence rate.

10 Sensor Movement based on Centralized Control The sensor needs to communicate with the base station. Report sensor ID, location, sensor type The deployment algorithm is carried out by the base station. Location mapping (assignment problem) The sensor movement is guided by the base station. Drawbacks: It is not suitable for large scale problem. The failure of base station leads to the failure of the whole system.

11 Swarm intelligence-based Movement What is SI-based algorithm? Stimulated from social insects society, like ants, bees, etc. simple rule carried by individual can lead to complex behavior of the whole system. TSP, network routing, optimization, etc. SI-based movement Each sensor carries a rule which guides the movement and leads to an optimal configuration of the sensor network. What’s the rule?

12 Rules carried by each sensor The movement direction is based on the current location and sensor type (where I am and where to go?) Rules carried by type 1 sensor: While x mod 2 !=0 and y mod 2 != 0 do if (x+y) mod 2 !=0 if x mod 2 == 1 then random walk N,S; else random walk W,E; end else random walk NW,NE,SW,SE; end

13 Performance metric Network Coverage Probabilistic sensing model: Probabilistic coverage: Convergence time The number of epochs of the last positioned sensor. Energy consumption Mechanical movement: Communication: g

14 Simulation and Evaluation… Simulation environment Sensing field: 50m×50m -> 8×8 grid points 64 sensors of 5 different types: r=1/4:1/4:1/4:1/8:1/8 Sensing distance: 10m Initial energy: Unit consumption in movement: Electric circuitry: Transmitter amplifier: Location of base station: (0,0)

15 Simulation and Evaluation… Convergence time

16 Simulation and Evaluation… Network coverage

17 Simulation and Evaluation… Energy consumption: randomSI-based

18 Simulation and Evaluation… Energy consumption:

19 Simulation and Evaluation… Energy consumption: Ratio =0.1 Ratio =0.446

20 Conclusion Optimal placement provides reliable coverage. Comparison of movement strategies. Random movement consumes more energy, no communication is needed. Centralized control converges the fast at the cost of full communication, where the energy consumption in communication is dominant. SI-based method provides a good tradeoff in terms of convergence time and energy consumption.