A Novel Particle Filter for Target Tracking in Wireless Sensor Network
Gang Lu,Wei Zhao,Jinping Sun,Shuqin Sun
DOI: https://doi.org/10.1049/cp.2013.0304
2013-01-01
Abstract:A novel method is presented in this paper, called modified converted measurements Kalman particle filter (M-CMK-PF), for target tracking in wireless sensor network (WSN). As an efficient improvement for particle filter (PF), this algorithm utilizes the modified converted measurements Kalman filter (M-CMKF) to estimate the posterior as an importance density for PF. The main idea of M-CMKF is converting polar measurements to Cartesian reference, calculating the converted error statistics and then performing the Kalman filter to obtain the posterior. Since there are no linearization errors of measurement model in the process, also the latest measurements are integrated with a prior, the M-CMKF generates importance density that approaches the real posterior more closely than the extended Kalman filter (EKF) and iteration extended Kalman filter (IEKF) which are filters in mixed coordinate. As a result, the M-CMK-PF has better tracking performance than the standard PF, EKF particle filter (EKF-PF) and IEKF particle filter (IEKF-PF). Additionally, the M-CMKF need not adjust parameters as the Unscented Kalman filter particle filter (UKF-PF) does, so the M-CMKPF is more robust in various applications. In addition, the calculation cost of the M-CMK-PF and EKF-PF are the smallest among the four. Simulation results demonstrated the effectiveness of our method.