Local decisions and optimal distributed detection in mobile wireless sensor networks

A set of mobile wireless sensors observe the environment as they move about and make decisions based on their observations. They send/relay their decisions to a sensor, called the Cluster-Head (CH), that has requested all decisions made about observations from a given region during a specified time-interval. There are two sources of error facing the multi-hop cluster of sensors that results from this scenario: observations are corrupted by noise and transmissions suffer communication errors. Once the sensors' decisions have reached the CH, the optimal maximum a posteriori (MAP) detector is known to be a weighted order statistic of these noisy decisions. We characterize the performance and energy usage of this decision fusion algorithm by: determining when local fusion reduces the CH's decision error rate and characterizing the tradeoff between the energy saved by compression of local decisions and the performance of the decision algorithm. Large deviation techniques, simulations and direct calculation are used to determine the performance of these strategies and to demonstrate that hybrids of them perform best.