面向地下电缆沟智能巡检机器人的紧耦合SLAM系统

Intelligent inspection robot for underground cable trench based on tightly coupled SLAM

摘要:

地下电缆沟的日常巡检劳动强度大，且存在危险隐患，是城市电力系统保持稳定工作亟需解决的问题，采用智能巡检机器人系统是解决这一问题的趋势。同步定位和实时地图的构建是地下电缆沟机器人自主巡检的前提。地下电缆沟等场景具有底纹理、结构化、路面平整度情况复杂、GPS信号差等场景特征，巡检机器人对该类结构化场景进行建图时，会出现地图退化和定位精度下降的现象。针对上述问题设计了一种基于多传感器的SLAM系统，融合了二维激光雷达、惯性测量单元、轮式里程计等多种传感器数据，通过自适应初始化对机器人里程计进行优化。针对不同路面平整度下的相邻激光关键帧匹配误差，设计了一种自适应帧间配准方法进行校正。现场试验表明，在路况复杂的地下电缆沟场景中，该方法比现有方法的地图退化率和定位误差平均分别降低了7.42%和8.73%，具有明显的工程应用价值。

Abstract:

The daily inspection of underground cable trenches is labor-intensive and has dangerous risks, which is an urgent problem to be solved to maintain the stability of urban power system. Intelligent inspection robot system is the trend to solve this problem. Synchronous positioning and real-time map construction capabilities are the prerequisite for autonomous inspection of underground cable trench robots. The underground cable trench has the characteristics of bottom texture, structure, complex road roughness, poor GPS signal and other scenes. The map degradation and positioning accuracy decrease often occur when the inspection robot builds the structured scene. To solve the above problems, a multi-sensor SLAM system is designed, which integrates the data of 2D LiDAR, inertial measurement unit and wheeled odometer, and optimizes the robot odometer through adaptive initialization. An adaptive inter-frame registration method is designed to correct the matching errors of adjacent laser key frames under different pavement flatness. Field tests show that the degradation rate and location error of the proposed method are reduced by 7.42% and 8.73% compared with the existing methods in the complex underground cable trench scenario.