Sensor scheduling and target tracking using expectation propagation

Multiple-sensor scheduling for target tracking applications using expectation propagation (EP) is examined. The method is an alternative to that of A.S. Chhetri et al. wherein an extended Kalman filter (EKF) was used to predict the next state for sensor scheduling purposes, and a sequential Monte Carlo particle filter (PF) method was used to implement the target tracking. In this application, EP is used instead of PF to estimate the unobserved state variable. Initial simulations show the EKF+EP (with scheduling) algorithm performs at least as well as EKF+PF, with a shorter run time and less programmatic complexity. EKF+EP (with scheduling) also performs better than EKF+EP (without scheduling)