Sequential Fusion with Truncated SPRT for Event Detection in Centralized and Decentralized Wireless Sensor Networks

In this work, the sequential fusion scheme is used to save the energy consumption for events detection in wireless sensor networks (WSNs). The fusion center (FC) sequentially accumulates the information forwarded by nodes, and broadcasts a termination message to stop the transmission of nodes when the accumulated information is enough for decision-making. Thus the average number of forwarded messages, as well as the average energy consumption, can be significantly reduced. Since the number of sensor nodes is generally limited, the truncated sequential probability ratio test (TSPRT) is applied at the FC. The main goal of this paper is to propose analytical approaches, for both centralized and decentralized detection, to determine the thresholds for TSPRT. According to the simulation results, with little performance degradation, the average number of transmissions can be greatly reduced compared with the nonsequential fusion scheme.