Fault-tolerant data aggregating using median filtering in cluster-based homogeneous WSN

The nodes of Wireless Sensor Network (WSN) easily produce fault sensor readings because of the harsh environment and the energy restriction and so on. To enhance reliability and save energy, this paper proposes a median filter and weighted pseudo-median filter based aggregating method for in-cluster data of homogeneous WSN, which combines fault nodes self-checking and clustering data fusion together. In the aggregating scheme, each node receives the fusion result from the cluster head (CH) and estimates the confidence grade of its measuring result for next aggregation round, if a node is not a fault one, it transmits its confidence grade and filtered measuring value to the CH. Based on the proposed weighting rule of median filter, the CH sets the current weight of each node´s data, which depends on the confidence grade determined by node itself and the current number of nodes to aggregate data in the cluster. The CH computes and outputs the weighted pseudo-median of the received data as the fusion result. The simulation and analysis show that on the premise of minimum redundancy, the proposed method is valid under the circumstance of high fault node rate and the high fusion accuracy is obtained even in the case of a small proportion of communication error or data packet dropout without communication detecting or confirmation.