Research On Key Technologies Of Intelligent Inspection Robot For Outdoor Substation Integrating Vision

The stable and safe operation of the substation ensures that hundreds of millions of people can safely and reliably use electric energy.The regular inspection of the substation is crucial to its safety.With the arrival of the era of artificial intelligence,manual inspection methods with many shortcomings such as low efficiency,low electromagnetic radiation,and missing detection errors are gradually disappearing from people’s vision.A more reliable and safe intelligent inspection robot has become an inevitable development of substation inspection technology.This paper presents a software and hardware system solution for an outdoor substation intelligent inspection robot.It mainly focuses on the four key technologies of robot odometer,map construction,autonomous navigation obstacle avoidance and pedestrian following,and tests the above technologies in the actual environment..The specific work is as follows:In terms of improving robot positioning accuracy.First of all,a variety of commonly used robot sensor models are given,including probabilistic motion models,pinhole camera models,and IMU sensor models.Then the calibration of multiple sensors is studied.For the calibration of sliding steering mobile platform and non-central installed lidar,a dual lidar auxiliary calibration method is studied.The camera and lidar sensor are jointly calibrated,and the verification experiment is given.In the Graph SLAM algorithm based on lidar.Firstly,the graph optimization theory and the Cartographer algorithm based on the graph optimization theory and the multi-sensor fusion odometer based on the unscented Kalman filter model are studied.Combining multi-sensor fusion odometer gives an improved Cartographer algorithm,In the indoor scene and large outdoor scene,the map construction experiment with multiple sensor configurations is given.Experimental results show that the improved algorithm can effectively improve the robot positioning accuracy and map construction accuracy.In terms of autonomous navigation and obstacle avoidance for intelligent inspection robots in substations.First,the system principle and composition of the open source navigation framework Navigation are studied,and then we conducted simulation experiments on a variety of algorithms including Dijkstra,GBFS,A * and TEB.Based on the experimental results,the A * algorithm for Manhattan distance and TEB algorithm is selected for global and local path planning.Then for the problem of the large obstacle blind area of the two-dimensional lidar,a depth camera is introduced,and a low obstacle detection and recognition algorithm is studied to improve the robot’s ability to actively avoid obstacles.The navigation,positioning and obstacle avoidance functions were tested including repeated positioning accuracy,narrow-track steering,and active obstacle avoidance.In terms of robotic pedestrian following.Firstly,the pedestrian detector YOLOv3-tiny was studied,and the lower body pedestrian detector was trained according to the needs of the subject.The FPS of the GPU-free device reached 27.42 fps/ s,which achieved multi-target detection,and then combined with Kalman filtering and Hungarian matching algorithm to achieve single target tracking,then combined with DWA dynamic window algorithm,a pedestrian tracking algorithm based on depth camera and lidar is studied,and finally the algorithm flow framework and pedestrian tracking experiments in single and multiplayer scenarios are given.Laid the foundation for many intelligent functions of substation inspection robots.