Adaptive Particle Filter with Estimation Windows

Abstract

Particle filter is well suited to estimate the state of non-linear non-Gaussian dynamic systems, which comes at the cost of higher computational complexity. But in many real time applications, it must deal with constraints imposed by limited computational resources. To deal with this question, we distribute the samples among the different observations arriving during a filter update, the novel algorithm represents densities over the state space by mixtures of sample sets. Another contribution of this paper is to increasing the efficiency of particle filters by adapting the size of sample sets during the estimation process. According to the relative entropy theory and particle number controller idea, we choose the number of samples, decrease computation overhead. A simulation of the classic HARD bearing only tracking problem is presented, the results show that the novel algorithm performs better than generic particle filter.