Research And System Design Of SLAM Algorithm For Power Inspection Robot

Point cloud map building algorithm is a key technology to the intelligent power station inspection robot which is equipped with a 2D Lidar.Point cloud map accuracy will directly affect the inspection robot's localization precision,and thus affect the inspection robot's update of movement state and the process of path planning.lt's the foundation for the inpection robot to fullfil autonomous navigation,and its importance is self-evident.The ICP algorithm is an algorithm commonly used in the process of building a point cloud map,but as the point cloud map built by the ICP algorithm takes longer to build and covers wider area,the cumulative error will become very serious.Loop clousure algorithm is an effective means to reduce the cumulative error,and has been widely studied by many scholars at home and abroad.One of the core problem in loop clousure is place recognition,whose target is to recognize the place which has been bypassed.An effective method to solve the problem of place recognition is to extract the feature points in the single frame data,and use the feature points to reflect the similarity between the two frames.Therefore,how to design the feature extraction algorithm for 2D Lidar and how to retrieve the similar frame by using the extracted feature has a definite value for solving the problem of place recognition.So the second chapter and the third chapter of this article are for these two issues.As for the first problem,considering the stable corner features such as building corner and table angle which are widely existed in the actual environment,a corner feature extraction algorithm based on 2D Lidar is proposed in this paper.Firstly,the algorithm combines the cosine distance between the two points in the point cloud and the cosine distance between the corresponding normal vectors to determine the neighborhood of each point,specifically,a larger Euclidean distance threshold is used to determine the rough neighborhood range,and a smaller cosine distance threshold is used to determine a more accurate neighborhood range.At the same time,in order to better extract the corners from the point cloud,this paper presents a novel evaluation function,which can effectively detect the exact corner.The comparison experiments on the online database show that the proposed algorithm is more accurate than other algorithms.As for the second problem,this paper propose a loop clousure algorithm based on the 2D Lidar corner feature.Firstly,we use the corner feature extraction algorithm proposed for the 2D Lidar in Chapter 2 to obtain the signature of single frame.Then we design a similarity comparison method to make the signature have rotation invariance property,and propose a method to calculate the relative pose of between the similar frames,then establish the graph model,and finally using the existing graph optimization framework to optimize the graph model.Experiments on the online public database show that the optimized cloud cloud map is better than the unoptimized point cloud map.Finally,in view of the map builiding and navigation project of electric power inspection robots cooperating with Dali technology company,this paper has developed a system combining map building,path planning and real-time navigation module,and applied the algorithm we researched to the system.The system has a very good practical effect.At present,the system has been accepted by the customer and delivered for use.