Research On Autonomous Visual Navigation Technology Of Filter Carrier In RCV

Nuclear power industry is the pillar industry in China.Vigorously developing nuclear power can optimize the energy structure and provide a large number of jobs,which is an important measure for our country.In order to ensure the normal operation of the nuclear power plant,its chemical volume system,RCV,requires to replace the filter regularly,but at present,manual transportation is still the only mode of transport.Due to the narrow road,lots of corners,inconvenient manual operation and the potential safety hazard of nuclear radiation,it is of great significance to study the autonomous navigation technology of the filter carrier vehicle and realize the autonomous replacement of the filter.Autonomous navigation is usually based on SLAM(simultaneous localization and mapping)algorithms.Lidar SLAM is the mainstream method because of its high ranging accuracy.However,there are many metal pipelines and plates in the working environment,and the lidar SLAM algorithm will lose a lot of data due to the reflection of the metal surface,then resulting in the failure of locating and mapping.However,visual SLAM is not influenced by metal environment,so it is significant to study the autonomous navigation technology based on visual SLAM.In order to solve the problems of data loss,failure locating on slope and ineffective observation of the surrounding scene in RCV environment with lidar SLAM,this paper proposed an autonomous binocular visual navigation method,including ORB-SLAM2 localization and mapping framework,path planning and path tracking algorithms.Considering the regularity of indoor slope,the visual feature points are truncated in the specified height interval and projected to a two-dimensional plane to establish a two-dimensional occupancy grid map.After processing,the path is planned and tracked.In order to solve the problem of too many turns in the original A* algorithm,an angle weighted A* algorithm of three adjacent nodes was proposed.And the OPEN list is implemented with the minimum heap,which makes the time complexity of node processing reach logarithmic order,so as to significantly reduce the path searching time.By introducing the angle weighting of three adjacent nodes into the heuristic function,the length and turns of the path are limited,so as to improve the performance of the algorithm.The improved A* algorithm is tested comparing with the original A* algorithm by changing the size of grid map and the proportion of obstacle nodes.The results indicate that the improved A* algorithm can effectively reduce the number of path turns,and the path length is close to the shortest.Finally,in order to verify the performance of the algorithms,establish the simulation and experimental slope environment to conduct key-frame poses tracking and reference path tracking experiments.The results indicate that the navigation scheme performs well in the slope environment under the condition of rich feature points.The slope boundary in the established two-dimensional grid map is obviously distinguishable,and it is easy to reprocess the map.The path planning algorithm can effectively plan the path and the absolute translation error in the map fulfill the accuracy requirements of the project.Therefore,The navigation system is feasible.