A Smart Sampling Particle Filter Algorithm For Global Localization Of Mobile Robots

Particle filter is a new real time inference algorithm which is based on Bayesian inference and Monte Carlo method. Because of its unique characteristics such as parallelizable, easy to implement, and efficient to process nonlinear problems, particle filter has been paid more and more attention in recent years and becomes a new and very promising hot topic in applied statistics, automatic control, and robotics. At the same time, the formidable increasing computational power and some recent developments in applied statistics and probability theory has stimulated many advances in this field. Although there are some theoretical and practical achievements, the use of particle filters is still in its infancy, there exist many basic issues need to be investigated.Traditional particle filter algorithm can lead to particle degeneracy and particle depletion problems in global localization of improvement strategies to increase particle filter's performance and its application to global localization of mobile robot. To improve the performance of particle filter, two techniques of particle degeneracy and particle depletion are studied. Comparing with previous methods, our algorithm can implement global locating of mobile robot with high precise.The main research work and dissertation are summarized as follows:Firstly, In this paper we proposed a novel improved particle filter algorithm named smart sampling particle filter (SSPF), there are three improvements as follows:1. In the phase of sampling, this algorithm generates a proposal distribution by UKF method and draws samples from it, and solves particle degeneracy caused by traditional particle filter using transition prior density function as proposal distribution.2. In the phase of resampling, this dissertation has studied the adaptive resampling algorithm. Different from the adaptive resampling algorithm based on effective sample size, a new adaptive resampling algorithm based on relative entropy is proposed which reduces traditional resampling step effectively. So this algorithm becomes more intelligence and can solve the problem of losing particles' diversity.3. Since one of the extreme conditions of resampling can cause particle depletion, we solve this problem by using Metropolis Hastings (MH) approach, which makes robot location more precise. By comparing with the simulation based on the real-coded PF algorithm, it has been confirmed that the proposed method has better performance.Secondly, this dissertation has studied the global localization of mobile robot which is basic but very important in independent intelligent robot field. So it's very challenge problem. This paper has studied SSPF algorithm for global localization of mobile robot. Simulation results show that SSPF can implement global locating of mobile robot with high precise.