Transformed space sampling particle filtering algorithm for single site passive localization

Passive localization from a single site is a typical nonlinear and non-Gaussian filtering and estimating problem, and usually suffers large initial estimation error and low observability. Considering the distribution of measurements is usually more peaked than the distribution of system state, an algorithm of transformed space sampling particle filter (TSSPF) is proposed, in which a new transformed sample space is constructed by using some auxiliary variables, thus the transformed space is a linear transform of observation space, and the proposal distribution can be obtained in the transformed space, from which transformed space particles can be sampled. The final target state particles are achieved by transforming the transformed space particles to target state space particles. Simulation results of comparing TSSPF with extend Kalman filter(EKF), unscented Kalman filter(UKF) and the EKF and UKF based hybrid particle filter, demonstrate that TSSPF is superior in convergence speed, tracking precision and filtering stability, and the estimation error can approximate the Cramer-Rao lower bound(CRLB).