Autonomous Exploration And Map Building Of Mobile Robot Based On Lidar/Camera Combination

Autonomous Exploration and mapping in unknown environment is one of the key technologies in the field of mobile robot.And it is also the premise for mobile robot to realize the tasks of search and rescue,reconnaissance,detection and detonation elimination in unknown environment.In the environment that human cannot or is not convenient to enter,the mobile robot needs to realize the active exploration of the unknown environment and build the global environment map through independent exploration and map building technology.According to the above requirements,this research focuses on the key technologies of autonomous exploration algorithm and 3D environment map construction of mobile robot in unknown environment.The main research contents are as follows:Firstly,aiming at the problem that local target points may be missed in traditional RRT search,an autonomous RRT search algorithm based on dynamic marginal constraints is proposed.With the help of the traditional idea of boundary exploration,the environment space is explored by the way of RRT search tree.The root node of the tree is used as the origin to divide the four quadrants,and the dynamic marginal constraint is imposed on the subtree,and then the efficiency of search is improved by the speediness and expansibility of the subtree.Secondly,aiming at the problem that the evaluation index of candidate exploration target point is too single and limited,the node cost index is designed.According to the passable path between the current tree node and the root node,the evaluation function of exploration target point is constructed by combining the navigation cost and information gain.According to the results of the evaluation function,the candidate target points are stored and updated in each round of exploration environment to determine the final exploration target points.Thirdly,aiming at the problem that LOAM algorithm can't eliminate the ground feature points,this paper improves the method of feature information extraction,and then extracts the edge feature points and plane feature points.In order to avoid clustering of feature points,the global environment is divided into regions,and the number of feature points in sub regions is specified.The improved ICP algorithm is used for feature matching.Fourth,aiming at the problem of blind area of vision in the initial position of lidar,this paper proposes to add auxiliary visual information in the way of static combination to make up for the problem of blind area of vision,and to realize the function of dynamic obstacle avoidance.Because the visual information is not directly added to the feature point cloud,it will not affect the real-time and accuracy of the map building algorithm.Fifth,the validity and practical effect of the algorithm are verified.Based on the lab's ROS based two wheel differential drive mobile robot platform,the prototype experiment is carried out,and the joint debugging of autonomous exploration and map building algorithm is carried out,and the experimental results of autonomous exploration algorithm and map building algorithm are analyzed respectively.