Research On Key Technology Of Substation Inspection Robot Based On 3D Laser

With the continuous construction and in-depth development of smart grids,traditional manual substation inspection methods have exposed shortcomings such as heavy workload,operation difficulty,and high risks.In order to meet the inspection requirements in substations,the demand for inspection robots in substations continues to increase.Among the related technologies of substation inspection robots,the positioning and navigation of the robot in the substation environment is one of the key technologies to realize the robot’s autonomous inspection.Due to the environmental requirements of the substation,the main sensor selected in the system is three-dimensional lidar,with IMU as an auxiliary positioning sensor.In order to realize the autonomous work of the inspection robot in the substation,the robot needs to have related functions.Firstly,the robot needs to build a high-precision three-dimensional point cloud map of the substation environment.Secondly,the three-dimensional point cloud map is used as a global map to realize its own global real-time positioning.Finally,the robot needs to go to the designated location to work according to the inspection tasks.According to the needs of the substation inspection robot in the three aspects of mapping,positioning and path planning,this paper constructs a high-availability system.The main research contents of this paper are as follows:Firstly,the algorithm of 3D map construction based on feature points is studied.In this paper,3D lidar is used to collect environmental data in the substation,and the curvature is used to eliminate the defects,extract and classify the feature points.The feature points can reflect the geometric characteristics of the environment more accurately which the robot in,reduce the influence of noise points on the matching algorithm.Constructing the constraint conditions of the point cloud matching algorithm with the feature points as the point set and correct the distortion of point cloud by uniform velocity model can make the pose estimation more accurate.At the same time,the constructed local map is used to modify the pose with low frequency to reduce the error of the final pose result.The experimental results show that the laser odometer can accurately restore the three-dimensional structural features of the substation and provide a high-precision map for robot positioning.Secondly,the lidar / IMU fusion localization algorithm based on global point cloud map is studied.The3 D lidar real-time point cloud data and the global large-scale point cloud map are matched by the normal distribution transformation registration algorithm to calculate the positioning data.The real-time point cloud from lidar is down sampled by the voxel grid filter,which greatly improves the calculation speed of the algorithm while ensuring the characteristics of the original point cloud and the positioning accuracy,so that the positioning system can meet the real-time updating requirements of the patrol robot for its own positioning.At the same time,the data of inertial sensor and lidar are fused by untracked Kalman filter algorithm to obtain the accurate and stable positioning data of the inspection robot.The high efficiency,reliability and accuracy of the algorithm are verified by the data collected from several actual scenes.Finally,the path planning algorithm based on fast-expanding random tree is studied.Based on the analysis of the classic fast-expanding random tree principle and simulation results,the existing algorithm is optimized by combining the bidirectional fast-expanding random tree and its improved algorithm in view of the shortcomings of the algorithm.Tract in the direction of generating new tree nodes in the random tree,traction includes the use of target probabilistic direct guidance and target indirect traction.This method reduces the sampling of the random tree in the direction away from the target point.By introducing the greedy strategy of RRT-Connect and simplifing the distance measurement algorithm,the random tree can search and expand in a feasible direction which speeds up the algorithm speed.After the algorithm successfully searched for a feasible path,the existing path is further optimized through node optimization,which greatly reduces the number of nodes and turning points in the path.The algorithm is simulated on a single obstacle map,a multiple obstacles map,and a map of the same scale as the substation site to verify the effectiveness of the algorithm.In order to realize the autonomous inspection of the substation,the system in this paper takes the industrial computer as the main control which carries the corresponding hardware modules.By using the ROS system as the software algorithm construction platform,it provides an effective solution for the mapping,positioning and path planning of the inspection robot.