The fault probability estimation and decision reliability improvement in WSNs

Faults are an essential fact in wireless sensor networking as coupled with a set of constraints. The decision making based on the reports of fragile and fallible sensor nodes might be very unreliable and, therefore, might fail to accomplish the tasks of WSNs. Previously, the reduction in the effects of faults is based on collaborative effort of a large number of sensor nodes. The collaborative work may consume valuable power and may fail as sensor nodes are severely affected by environmental interference. In this paper, instead of the neighborhood communication and pure data fusion, the collaborative effort in decision making is accomplished based on the fault probability estimation, reading variance estimation, and critical value adjustment. The fault probability of sensor nodes is computed by their reports and the t-out-of-n rule to make reliable decisions. The reading variance is estimated by well-known sample variance to assess the effects of environmental interference. The critical value adjustment is triggered as estimated reading variance changed, high fault probability, or decision quality unsatisfied to reduce the bias of fault probability.