Cooperative control, learning and sensing in mobile sensor networks

Abstract

Scope and Method of Study:

Mobile sensor networks (MSNs) have great potential in many applications including environment exploring and monitoring; search and rescue; cooperative detection of toxic chemicals, etc. Motivated by the broad and important applications of MSNs and inspired by the cooperative ability and the intelligence of fish schools and bird flocks, this dissertation develops cooperative control, learning and sensing algorithms in a distributed fashion for MSNs to realize coordinated motion control and intelligent situational awareness.

Findings and Conclusions:

The proposed algorithms can allow MSNs to track a moving target efficiently in cluttered environments and even when only a very small subset of the sensor nodes know the information of the target; adjust their size (shrink/recover) in order to adapt to complex environments while maintaining the network connectivity and topology; form a lattice structure and maintain the cohesion even when the measurements are corrupted by noise; track multiple moving targets simultaneously and efficiently in a dynamic fashion; learn to evade the enemy (predators) in a distributed fashion while maintaining the network connectivity and topology; estimate and build the map of a scalar field. We conducted several experiments using both simulation and real mobile robots to show the effectiveness of the proposed algorithms. We also extended our framework to cooperative and active sensing in which the mobile sensors have the ability to adjust their movements to adapt to the environments in order to improve the sensing performance in a distributed fashion.