基于MCMC粒子滤波的机器人定位

Robot localization based on MCMC particle filter

针对基于传统粒子滤波的机器人定位方法存在粒子退化的问题,提出了基于马尔科夫蒙特卡罗(MCMC)粒子滤波的机器人定位方法.将传统的粒子滤波算法与典型的MCMC方法——Metropolis Hastings(MH)抽样算法有机结合,并应用于机器人定位研究中.试验结果表明,MCMC方法可以有效抑制粒子退化问题.与基于传统粒子滤波的机器人定位方法相比,该方法降低了定位误差均值和定位误差最大值,取得了更高的定位精度,有效地解决了机器人定位这一非线性非高斯状态估计问题.

Robot localization based on the Markov Chain Monte Carlo （MCMC） particle filter was proposed to solve the problem that robot localization based on the simple particle filter suffers from severe sample degeneracy. The standard MCMC method, Metropolis Hastings （MH） sampling, was incorporated into the filtering framework, and was applied to the robot localization problem. Experimental results showed that the MCMC particle filter can increase the sample variety and reduce sample degeneracy. Robot localization based on the MCMC particle filter is much more accurate, compared with robot localization based on the simple particle filter.