WIRELESS SENSOR NETWORK AUTONOMIC RELIABILITY EVALUATION BASED ON COARSENING COMBINATION EVIDENCE REASONING

摘要

As for the extremely unpredictable factors of Distributed Sensor Networks (DSNs) with constrained resources operating in an unattended mode in uncertain dynamic environments, the behavior evaluation of sensor-network nodes is vital to architecture of autonomic and fault-tolerant DSNs. Because of Bayesian probability incapability of capturing epistemic uncertainty (uncertainty with little prior-knowledge), one evaluation scheme based on Dempster Evidence Theory (DST) is proposed with coarsening and refining combining algorithm by Fast Moebius Transform (FMT). This method offers an efficient framework for uncertainty quantifying and partial knowledge processing with sharply decreasing computation complexity. Simulation verified this ubiquitous low-cost computation appropriate to flexible Wireless Sensor Network Management Protocol (SNMP) for prolonging live-time of sensor network.