Agent-based fault detection mechanism in wireless sensor networks

Recently, agent-based approaches are considered as an appropriate solution to address the issue of the overwhelming data traffic in Wireless Sensor Networks (WSNs). Theoretically, these approaches would eliminate the redundancy and achieve substantial energy gain. However, in practice, the reliability of sensor devices has been recognized as one of the crucial issues in wireless sensor networks. Using agents at the sensor devices may provide more efficient energy consumption. But, micro-sensors are subject to high-frequency faults in distributed environments. Towards this end, we propose agent-based system architecture with fault-detection inference engine based on reverse multicast tree to evaluate sensor nodes' fault probabilities. Due to the characteristics of wireless sensor networks (energy awareness, constraint bandwidth and so on); it is infeasible to require each sensor to announce its working state to a centralized terminal node. Therefore, we formulate the agent's inference engine as nondeterministic finite accepter and adopt iterative computation to infer the fault probability of nodes in reverse multicast tree.