Abstract
This paper addresses schemes for fault detection and isolation in a semi-decentralized environment. Now-a-days, sensor fault and failure are prevalent issues in numerous wireless sensor networks. We propose a few algorithms based on simple phenomenon of data fusion. Initially, a mutual consensus has been built among followers (e.g., Unmanned Autonomous Vehicles in this case) who are tracking a combine target. Having known the followers, relative positions with respect to target, a median is computed by each follower. This median is then shared with immediate and extended neighbours to compare with their estimated values about the same target position. If estimation is beyond the prescribed limits, the follower (sensor) is diagnosed as faulty, otherwise is considered healthy. Three different types of induced faults are discussed here: (i) follower – target or line of communication fault; (ii) follower – follower or communication with neighbour fault; and (iii) simultaneously these two faults. The scenario wherein eight followers are tracking a combine target in circular fashion has been considered to elaborate these faults
