基于视觉的配电房巡检机器人避障方法

Obstacle avoidance method of power distribution room inspection robot based on vision

摘要

针对配电房巡检机器人在复杂动态环境中实现高效自主导航与实时避障的需求,提出一种基于机器视觉的动态避障方法。该方法首先融合改进的YOLO-v8目标检测算法,对环境中的静态与动态障碍物进行实时识别与位置估计,从而克服传统移动机器人避障算法在复杂场景下识别精度低、响应迟滞等问题。其次,结合动态窗口法(dynamic window approach,DWA)进行局部路径规划优化,将YOLO-v8检测得到的障碍物空间坐标及运动参数动态映射到DWA速度空间中,实现速度与方向的实时调整,增强机器人在动态环境中的避障与路径稳定性。最后,在仿真环境中对算法进行测试与验证。实验结果表明,该方法能够在复杂动态场景下准确识别并避让多种障碍物,提升了巡检机器人的自主导航安全性与智能化水平。

Abstract

To meet the demand for efficient autonomous navigation and real-time obstacle avoidance of inspection robots in complex dynamic power distribution rooms, this paper proposes a vision-based dynamic obstacle avoidance method. Firstly, an improved YOLO-v8 algorithm is applied for real-time detection and localization of static and dynamic obstacles, addressing the limitations of traditional mobile robot avoidance algorithms in accuracy and responsiveness. Secondly, the dynamic window approach (DWA) is integrated for local path planning optimization, where obstacle position and motion parameters obtained from YOLO-v8 are mapped to the DWA velocity space in real time, enabling adaptive speed and direction control. Finally, simulation experiments are conducted to verify the proposed method. Results show that the algorithm can accurately detect and avoid various obstacles in dynamic environments, improving the safety and intelligence of inspection robots′ autonomous navigation.

关键词

机器视觉 / 目标检测 / YOLO-v8 / 动态窗口方法 / 局部路径规划 / 动态避障

Key words

machine vision / object detection / YOLO-v8 / dynamic window approach / local path planning / dynamic obstacle avoidance