Mobile tracking and parameter learning in unknown non-line-of-sight conditions

This paper studies the mobile tracking problem in mixed line-of-sight (LOS) and non-line-of-sight (NLOS) conditions, where the statistics of NLOS error is Gaussian with fixed but unknown mean and variance. A Rao-Blackwellized particle filtering and parameter learning method (RBPF-PL) is proposed, in which the particle filtering with optimal trial distribution is first applied to estimate the posterior density of sight conditions, then the decentralized extended Kalman filter (EKF) is used to estimate the mobile state. In the parameter learning step, using the conjugate prior distribution on the unknown parameters, the posterior distribution of unknown parameters is further updated according to the sufficient statistics. Simulation results show the RBPF-PL method is effective to infer the unknown NLOS parameter and could achieve good tracking performance using small number of particles.