论文部分内容阅读
在标准FastSLAM中,随着重采样次数的增加会出现十分严重的粒子退化现象,从而导致机器人位姿估计的一致性很差.针对FastSLAM算法的这一缺陷,提出一种改进的FastSLAM算法.此算法在标准FastSLAM的重采样条件判断中,额外考虑了粒子权重协方差和每个粒子的测量残余一致性,并且使用指数等级选择算法进行新粒子的生成.通过仿真实验可以看出,改进的FastSLAM算法不但可以明显地提高机器人位姿估计的一致性,而且能够很好地保持粒子多样性.
In the standard FastSLAM, a very serious particle degradation occurs with the increase of the number of resampling, which leads to the poor consistency of robot pose estimation.An improved FastSLAM algorithm is proposed based on the drawback of FastSLAM algorithm. In the judgment of re-sampling conditions of standard FastSLAM, the particle weight covariance and the measurement residual consistency of each particle are additionally considered, and the generation of new particles is performed by exponential level selection algorithm.It can be seen through the simulation experiment that the improved FastSLAM algorithm Not only can the consistency of robot pose estimation be significantly improved, but also the particle diversity can be well maintained.