Energy Efficient Decentralized Detection

In Wireless Sensor Networks (WSNs), event detection problem has been formulated as a binary hypothesis testing problem, and past work is largely focused on problem formulations that assume sensor nodes in a parallel configuration, where individual hard or soft decision is computed at each sensor node and transmitted directly to a fusion node. In such a configuration, sensor nodes farther away from the fusion node use more power to transmit their decisions. In this paper, we investigate a hierarchical configuration of the sensor nodes. In our proposed scheme, each sensor's decision is made aiming at minimizing the probability of error in the fusion node while imposing constraints on the energy consumption for information transmission. The solution is based on optimally choosing the bit allocation among the sensors and the thresholds of the decision rules. Simulation results show significant improvement in the case of the proposed hierarchical configuration compared with parallel configuration for different size networks, especially in larger networks. For example, in a 33-node network, for a fixed total number of transmitted bits, 71% less energy is consumed, and for a fixed energy budget, 146% more bits are transmitted, and 46% higher computed Chernoff information is available at the fusion node.