基于多模态感知的敬老院智能机器人SLAM算法研究

打开文本图片集

关键词：多模态融合；位姿图优化;RTAB-Map;RGB-D;激光雷达

中图分类号：TP242 文献标志码：A 文章编号：1003-5168（2025）24-0022-06

DOI： 10.19968/j.cnki.hnkj.1003-5168.2025.24.004

Research on SLAM for Intelligent Robots in Gerocomiums Based on Multimodal Sensing

WANG Xiaotao ZHANG Hongyan LIU Huijie ZHANG Haitao SUN Renjie （JilinUniversityofChemical Technology，Jilin132O11，China）

Abstract： [Purposes] This study aims to solve the problems of edge misalignment，mapping blanks，and inconsistent global maps in the SLAM mapping of intelligent robots in gerocomiums.[Methods] By researching and comparing the Gmapping， ORB-SLAM2， and RTAB-Map mapping algorithms， we propose an optimized RTAB-Map algorithm.The optimization incorporates the global pose graph optimization mechanism of GTSAM，nonlinear least-squares constraint solving，and an Extended Kalman Filter （EKF）multi-sensor fusion architecture. The optimized RTAB-Map algorithm is then used to perform field mapping in real environments by fusing data from three sensors： a single-line LiDAR，an IMU，and an RGB-D camera.[Findings] Experimental results show that the robot's mapping trajectory drift rate is reduced by 68% ，the robot's state estimation error is decreased by 42% ，the edge alignment error in narrow indoor spaces is ⩽5cm ，the map completeness in low-light scenariosis increased by 89 % ，and the system stability under dynamic interference is improved.Furthermore，thefiltering performance for slowmoving elderly individuals （dynamic objects） during mapping is improved by 50% .[Conclusions] The improved RTAB-Map algorithm cansolve the problems of edge misalignment，mapping blanks，and inconsistent global maps encountered by robots during SLAM mapping in gerocomiums，providing technical support for the application of robots in such settings.

Keywords：multimodal fusion; pose graph optimization;RTAB-Map;RGB-D; LiDAR

0 引言

目前，敬老院智能机器人面临的难题是智能机器人在室内狭小的空间建图会出现边缘错位，在强光或者弱光环境下会丢失环境纹理信息和环境深度信息，导致出现建图空白、里程计累计误差等问题，使得机器人位姿估计漂移与地图全局不一致。（剩余5686字）