基于一致性的FastSLAM算法的优化  

Optimization of FastSLAM Algorithm Based on Consistency

作　　者：耿小毛 

机构地区：[1]上海理工大学理学院,上海

出　　处：《理论数学》2024年第1期302-317,共16页Pure Mathematics

摘　　要：本文从可观测性的角度研究FastSLAM算法的一致性问题并对算法进行了改进。相比较于传统的FastSLAM算法,一种改进的FastSLAM算法被提出从而获得更好的一致性性能。首先,在重要性采样阶段,对象标记法用于清晰地标注单个粒子状态。此外,在地图估计阶段,将第一估计雅可比矩阵(FEJ)与扩展卡尔曼滤波相结合以此提高了算法的一致性。最后,通过仿真实例验证了改进的FastSLAM算法的有效性。In this paper, the consistency of FastSLAM algorithm is improved from the perspective of observability. Rather than the traditional ones, an improved FastSLAM algorithm is proposed to achieve better consistent performance of FastSLAM. First, in the importance sampling phase, the object labeling method is used to clearly label individual particle states. Moreover, in the stage of map estimation, the First Estimates Jacobian (FEJ) is combined with the extended Kalman filtering to improve the consistency of the algorithm. Finally, the effectiveness of the proposed improved FastSLAM algorithm is validated through a simulation example.

关 键 词：移动机器人 扩展卡尔曼滤波 FASTSLAM Kullback-Leibler散度(KLD)采样 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]