基于改进的粒子滤波蒙特卡洛定位算法研究

Research on Monte Carlo Localization Algorithm Based on Improved Particle Filter

在未知先验环境下,移动机器人的自身定位是自主导航的核心和基础。基于粒子滤波的MCL算法存在着粒子退化、处理能力方面难以达到实时要求以及计算量较大等问题。提出一种基于改进的粒子滤波MCL算法,在SR-CKF设计的提议分布基础上添加EKF扩展为多提议分布粒子滤波算法。为解决因忽略当前时刻观测信息导致的粒子退化问题,在设计的提议分布中,加入系统当前时刻的观测信息,并以相对的比率对粒子进行采集,使粒子收敛于观测似然较高的区域中。利用KLD重采样来估计当前时刻粒子在系统下状态空间的分布情况,并通过在线调整机制对下一时刻粒子数目进行调整,以达到减小计算量的目的。根据Matlab仿真对比实验可以看出,改进后算法的均方根误差减小至1.947 cm,运行时间缩短了29.7%,有效的粒子百分比达到82.6%,平均粒子数为70.47,减弱了粒子退化问题从而达到精准定位的效果。最后通过ROS机器人开源操作系统对算法的有效性和可行性进行进一步验证分析。

In the unknown a prior environment,the mobile robot's own localizaiton is the core and foundation of autonomous navigation.The MCL algorithm based on particle filter has problems such as particle degradation,difficult real-time requirements in processing ability and large computational complexity,an improved particle filter MCL algorithm is proposed based on the proposed distribution of the SR-CKF design,EKF is added to expand the multi-proposed distributed particle filter algorithm.In order to solve the problem of particle degradation caused by ignoring the current observation information,in the proposed distribution of the design,the observation information of the current moment of the system is added,and the particles are collected at a relative ratio,so that the particles converge to the region with high observation likelihood.KLD resampling is used to estimate the distribution of particles in the state space under the system at the current moment,and the number of particles at the next moment is adjusted by an online adjustment mechanism to reduce the amount of calculation.According to the Matlab simulation experiment,after the improved algorithm,the root-mean-square error was reduced to 1.947 cm;the running time was shortened by 29.7%;the effective particle percentage was 82.6%;and the average particle number was 70.47,which reduced the particle degradation problem and achieved the effect of accurate locating.Finally,the effectiveness and feasibility of the algorithm are further verified by the ROS robot open source operating system.