摘要
利用学习与进化结合的思想,改善基于粒子滤波的SLAM算法。在对学习与进化的关系深入分析的基础上,针对基于粒子滤波的SLAM算法,提出将滤波过程分成学习和进化两个阶段,分别给出相应算法解决粒子有效性与多样性的问题,缓解二者之间的矛盾,改善了SLAM算法的效果,增强了算法的鲁棒性,也验证了学习与进化的关系。最后,通过多次Monte-Carlo仿真实验结果表明了该算法的有效性。
The main contribution is utilizing both learning and evolution method to improve the performance of SLAM algorithm based on particle filter.The particle filter was considered two parts:first part played the learning role and the another one played the evolution.These two parts could be used to solve the sample impoverishment problem and the degeneracy problem for particle filter respectively.In such case,the filter was more robust and performs better.For this purpose,different algorithms for each part were proposed.In the mean time,the relationship between learning and evolution was proved again.Finally,the result of Monte-Carlo simulation proves the algorithm is valid.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2010年第5期1204-1209,共6页
Journal of System Simulation
基金
国家863项目资助计划(2006AA040203)
国家863项目资助计划(2009AA04Z213)
国家自然科学基金(60875057)
关键词
学习
进化
粒子滤波器
同时定位与地图创建
learning
evolution
particle filter
simultaneous localization and mapping (SLAM)