Energy-efficient Estimation And Filtering In Wireless Sensor Networks

In a distributed wireless sensor network (WSN), local sensors cooperate with a fusion center (FC) in estimating an unknown parameter or tracking a dynamic state in the environment. The local sensors are in charge of observation and sending the compressed data to the FC, and the FC aggregates the data and generates a final estimation. Due to the specific application environment, sensor networks face serious energy and bandwidth constrains. Thus we must design the best estimation or filtering scheme under the limited energy and bandwidth.In this paper, we propose a variable-length Time Division Multiple Access (TDMA) scheme and use the uncoded MQAM as the modulation scheme for all the sensors. We first investigate the problem of energy-efficient distributed estimation in wireless sensor networks. In an inhomogeneous sensing and transmission environment, the minimization of total energy is jointly determined by the optimal quantization and transmission scheduling. In order to minimize the total energy under the mean squared error (MSE) and total transmission time constraints, we solute our problem with 3 steps. We first deduce a closed-expression of optimal quantization with transmission time fixed via convex optimization. Then we work out the optimal transmission time scheduling via MoveRight algorithm when the quantization lengths are fixed. Based on the first two steps, we design a joint algorithm which iteratively computes the optimal quantization lengths and transmission times. We proved that the algorithm is convergent. Simulations show that the iteration converges quickly, and significant energy saving can be achieved when compared with the uniform quantization and transmission scheme. The second problem we investigate in this paper is the distributed Kalman filtering based on quantized innovations. We first work out a closed-expression of our new Kalman filter with quantized innovations. Then we discuss the optimal quantization length for every sensor, in order to get the best MSE performance under a given energy consumption. The simulation results show that our quantization scheme get a significant transmission energy saving under the inhomogeneous transmission environment.