Adaptive retransmission with balanced load for fault-tolerant distributed detection in wireless sensor networks

An adaptive retransmission approach recently has been proposed to improve the misclassification probability of dis-tributed detection with error-correcting codes in wireless sensor networks. A code matrix is designed and a code-word is assigned to each hypothesis. Each sensor is asso-ciated with a decision pattern and has a set of thresholds to make its local decision on its detection result. However, when some sensors have unrecognized faults, a fusion cen-ter tends to ask the sensor with the same decision pattern as the faulty sensor to retransmit its decision. This tendency causes unbalanced load of the network. This work proposes a novel adaptive retransmission algorithm with balanced load to combat this problem. Each sensor carries all sets of thresholds. When the decision based on a set of thresholds must be retransmitted, a sensor is randomly selected. The selected sensor then makes its new decision according the threshold and its detection result. Because of the random selection, the load is balanced. Moreover, computer sim-ulations show that the misclassification probability of the load-balanced algorithm is close to the previous approach.