基于序列Monte Carlo方法的运动目标跟踪

针对移动机器人中的非线性非高斯特性,提出用粒子滤波的方法对运动目标进行跟踪,研究在贝叶斯框架下进行目标跟踪的原理,分析粒子滤波算法及存在的问题.应用这种方法,针对移动机器人的实际情况,建立了具体的系统模型和测量模型,比较不同的重采样算法,实现了对运动目标的跟踪任务.仿真结果表明,粒子滤波对于非线性非高斯的动态系统有良好的估计效果,在3种常用的重采样算法中,残差重采样的准确性稍高,分层重采样的计算比较简单,而多项式重采样是一种基本的重采样算法.

Sequential Monte Carlo methods for moving target tracking

In light of the mobile robotics' characters of nonlinearity and non-Gaussian,the method of particle filter for moving target tracking is presented.The principle of target tracking is studied based on the Bayesian framework; and the algorithm of particle filter and its problems are analyzed. Using this approach,concrete system model and measurement model are built for mobile robots, also different resampling algorithms are compared. Finally, we finished the task of moving target tracking. The simulation experiments show that the particle filter has perfect estimating results for nonlinear/non-Gaussian dynamic system;what is more, among these three common resampling algorithms, the residual resampling is more accurate; the stratified resampling has less computation; and the multinomial resampling is an essential method.