Scholarlink

<p>In order to realize high-precision mapping of mobile robots in unknown indoor environments, this paper compares and analyzes the Gmapping mapping algorithm and the Hector mapping algorithm in a simulation environment. On this basis, the Gmapping algorithm is optimized in the backend to solve the particle degradation problem caused by large-angle turning or rapid environmental changes. The particle weights are processed in the laser scanning part, and the Gaussian function is used to smooth the weights of abnormal particles with large weight differences. Then, the resampling strategy is optimized, and a low-variance resampling strategy is used to reduce particle degradation and further improve particle diversity. Finally, the number of particles is adaptively adjusted to improve the efficiency of calculation. The simulation and physical experimental results show that the root mean square error of the absolute error of the improved algorithm in this paper is reduced by 39.8% and the standard deviation is reduced by 71.5% compared with the original algorithm in the estimated pose and true pose in the mapping. The relative error is also significantly reduced.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on mobile robot mapping based on improved Gmapping algorithm

  • 期刊论文

  • 发表日期：

    2025年12月5日

查看全文
  • Jing Ma,
  • Tong Zhang,
  • Meng Wang,
  • Shan Lu,
  • Chuanlong Li,
  • Jun Xu

摘要

In order to realize high-precision mapping of mobile robots in unknown indoor environments, this paper compares and analyzes the Gmapping mapping algorithm and the Hector mapping algorithm in a simulation environment. On this basis, the Gmapping algorithm is optimized in the backend to solve the particle degradation problem caused by large-angle turning or rapid environmental changes. The particle weights are processed in the laser scanning part, and the Gaussian function is used to smooth the weights of abnormal particles with large weight differences. Then, the resampling strategy is optimized, and a low-variance resampling strategy is used to reduce particle degradation and further improve particle diversity. Finally, the number of particles is adaptively adjusted to improve the efficiency of calculation. The simulation and physical experimental results show that the root mean square error of the absolute error of the improved algorithm in this paper is reduced by 39.8% and the standard deviation is reduced by 71.5% compared with the original algorithm in the estimated pose and true pose in the mapping. The relative error is also significantly reduced.