Research On Path Planning Method Of Mobile Robot In Substation Environment

In order to improve the inspection efficiency of substations,reduce safety risks and labor intensity of workers,using intelligent inspection robots instead of manual inspections has become an effective means of achieving automated inspection,detection,and analysis of substation equipment status.In substation inspection,adopting a mobile robot solution based on cloud devices is an efficient choice.However,this solution needs to consider the impact of network transmission on time overhead,as well as the high deployment and maintenance costs of cloud devices.In contrast,offline mobile robots exhibit better real-time performance in path planning,but with limited computing power and storage space,there are issues with low real-time planning of mobile paths and slow obstacle detection speed.In response to these issues,this article has conducted the following research in order to better achieve autonomous inspection of substation inspection robots in offline state.Firstly,in response to the problems of low efficiency and multiple search nodes in global path planning using the A * algorithm for offline inspection machines in substations,this article makes the following improvements:(1)optimize the heuristic function to more accurately estimate the distance cost between two points,thereby improving the convergence speed of the algorithm,reducing the number of traversal nodes and time overhead;(2)Implement constraints on the scope of search nodes,reduce the number of reverse search nodes,improve search efficiency,and add strategies to remove redundant nodes,obtain key nodes,and reduce storage space occupation;(3)Use B-spline method to smooth the path and improve its feasibility.Through simulation experiments,compared to traditional A * algorithms,the improved A * algorithm in this article reduces time overhead by 60%,reduces the number of search nodes by 85%,significantly reduces the number of generated path nodes,and the path inflection points are smoother.Secondly,in response to the problem of high computational complexity and memory consumption when using the YOLOv5 model for real-time obstacle detection during the movement of offline mobile robots in substations,this article has made improvements to the YOLOv5-6.0 version:(1)using the improved lightweight network PP-LCNet as the feature extraction backbone,replacing some CBS modules in Neck with the Depth Sep Conv module in PP-LCNet,which reduces model parameters and improves inference speed;(2)Use Soft-NMS for subsequent processing to improve model accuracy and reduce missed detections.After simulation experiments,the obstacle detection model in this article has reduced the parameter count by 38%,computational complexity by 45%,and average accuracy slightly compared to the YOLOv5 s model.At the same time,FPS has slightly improved,providing strong support for real-time path planning and adjustment.Finally,in response to the issue of how to safely and effectively avoid obstacles detected by offline mobile robots in substations during their movement,this paper proposes a quadratic path planning method in the substation environment.Key points are obtained from the global path planning algorithm as local target points for the dynamic window method,in order to avoid the dynamic window method falling into local optima.When obstacles are detected,The dynamic window method based on physical modeling of robots can plan reasonable routes suitable for robot movement.In the simulation experiment,this article not only verified the substation environment.