Fusion of Local Decisions based on Rao test in Resource-constrained Sensor Networks

Detection is a basic task of a wireless sensor network (WSN). To meet severe bandwidth and energy limitations of WSNs, network nodes are usually programmed to decide locally about a desired event occurrence and send just one bit to a fusion center (FC) wherein an optimum decision must be taken. In this paper, we employ the Rao test for fusing local decisions of nodes. The Rao test is well-known for its much lower computational complexity than the more common (but computationally heavier) generalized likelihood ratio test (GLRT). We obtain the closed forms of the Rao-test-based fusion rules in both homogeneous and heterogeneous WSNs in the presence of imperfect communication channels. The importance of the presented formulations lies in their simplicity while taking practical issues into account. We show that the Rao-test-based decision fusion in homogeneous WSNs coincides with the counting rule (CR). Moreover, simple methods of adjusting local detection thresholds are proposed. The effectiveness of RAPID in the improvement of the overall network performance is shown through simulations in different scenarios.