结合传感器融合与退火参数改进RBPF-SLAM算法  

Improving RBPF-SLAM Algorithm with Sensor Fusion and Annealing Parameters

作　　者：李旦 王冠凌 丰宇航[1,2] LI Dan;WANG Guanling;FENG Yuhang(School of Electrical Engineering,Ministry of Education,Anhui Polytechnic University,Wuhu Anhui 241000,China;Key Laboratory of Advanced Perception and Intelligent Control of High-end Equipment,Ministry of Education,Anhui Polytechnic University,Wuhu Anhui 241000,China)

机构地区：[1]安徽工程大学电气工程学院,安徽芜湖241000 [2]安徽工程大学高端装备先进感知与智能控制教育部重点实验室,安徽芜湖241000

出　　处：《海南热带海洋学院学报》2021年第2期81-87,共7页Journal of Hainan Tropical Ocean University

基　　金：皖江高端装备制造协同创新中心开放基金项目(GCKJ2018007)。

摘　　要：Rao-Blackwellized粒子滤波器(RBPF)可以有效地处理同时定位和地图构建(SLAM)的问题。传统的RBPF-SLAM算法中因使用的粒子数目多和频繁执行重采样,导致构建的栅格地图精度不高。针对算法中存在的问题,提出了一种改进的RBPF-SLAM算法。通过扩展卡尔曼滤波融合轮式里程计和IMU的数据提高移动机器人的位姿精度,由融合后的里程计运动模型与激光雷达预测模型作为混合提议分布,利用退火参数优化提议分布减少采样所需粒子数目,并引入自适应重采样减少重采样的次数。由仿真实验结果表明:改进的RBPF-SLAM算法中使用15个采样粒子比传统算法使用30个采样粒子构建二维栅格地图效果更好,且构建地图的运行时间缩短约28%。Rao-Blackwellized Particle Filter(RBPF)can effectively deal with simultaneous localization and mapping(SLAM)issues.The traditional RBPF-SLAM algorithm uses a large number of particles and frequently performs resampling,resulting in low accuracy of the constructed raster map.Aiming at the problems in the algorithm,an improved RBPF-SLAM algorithm was put forward.To improve the pose accuracy of the mobile robot,the extended Kalman filter was applied to fuse wheeled odometer and IMU data.The fused odometer motion model and the lidar observation model were used as the mixed proposal distribution,and the annealing parameter was used to optimize the proposal distribution to reduce the number of particles required.Meanwhile,adaptive resampling was introduced to lessen resampling.The simulation results indicated that the improved RBPF-SLAM algorithm which used 15 sampled particles is better at building a two-dimensional grid map than the traditional algorithm which uses 30 sampled particles,resulting in reducing 28%of the running time of building the map.

关 键 词：RAO-BLACKWELLIZED粒子滤波器 SLAM 传感器融合 退火参数 重采样 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]