Local information guided autonomous exploration in sensor networks: Algorithms and experiments

This paper presents simple and effective technologies that support autonomous robotic exploration in wireless sensor networks where neither location information nor global network knowledge is known by the robot or sensors. Under our proposed approach, the sensor network is divided into convex partitions. In each partition, a small constant number of sensor nodes are identified as landmarks to establish a virtual coordinates system. Then, the robot employs a progressive refinement algorithm based on inaccurate virtual coordinates to discover the targets. Both the coordinates establishment and the robotic exploration are based on local information obtained in sensor networks. The algorithms are light-weight, with very limited computation and communication overhead introduced to the robot and sensors. The proposed approach is prototyped and experimentally evaluated by using an iRobot Create® Programmable robot and 36 Crossbow MICAz motes, providing useful empiric insights under realistic environments. Simulations are further carried out to study its performance in large scale networks.