Autonomous Navigation Of Mobile Robot In Unknown Dynamic Environment

With the development of intelligent perception and intelligent control technology,the autonomous ability of robots has also been improved,and widely used in daily production,intelligent logistics,rescue and other fields.The autonomous navigation capability of an intelligent robot is one of the key technologies of its intelligence.Most of the existing navigation methods are based on predetermined scenarios.When facing unknown dynamic scenarios and real-time moving dynamic obstacles,the robot cannot It can effectively realize dynamic unknown scene mapping and dynamic obstacle avoidance.Therefore,this thesis uses lidar as the source of scene information perception,analyzes the problems existing in dynamic obstacle detection,point cloud map construction,dynamic scene description and path planning in unknown dynamic environments,and develops autonomous navigation methods for mobile robots in unknown dynamic environments.Research and verify it in the experimental platform.The research content of this thesis is mainly aimed at the unknown dynamic scene,the autonomous navigation research is carried out in the scene where the map cannot be constructed in advance and there are real-time dynamic obstacles.The main research contents are as follows:(1)In order to solve the detection problem of dynamic obstacles in an unknown dynamic environment,this thesis tracks the trajectory of each obstacle by fusing multiframe point cloud data,and determines whether the object is a dynamic obstacle.Firstly,the pose estimation algorithm based on nonlinear optimization is used to obtain the pose information of the robot according to the characteristics of the point cloud;secondly,the original point cloud is subjected to ground removal,clustering and projection processing;Finally,IoU-tracker based on point cloud is used for multi-target tracking,static and dynamic obstacles are judged according to the historical trajectory information of obstacles.The average running time of this method on the experimental platform is 65 ms,which meets the real-time requirement.(2)In order to eliminate the influence of dynamic obstacles on static point cloud mapping and solve the problem of scene description in dynamic environment,a method of constructing static point cloud map in dynamic environment based on real-time detection and a method of describing dynamic and complex scene based on hierarchical cost map are proposed.Firstly,according to the detection results of dynamic obstacles,in the process of constructing the point cloud map,the point cloud on the dynamic obstacle is removed.Then,the calculation amount of updating the map is reduced by layering the cost map.Finally,compared with the traditional mapping algorithm,the algorithm in this thesis can construct a static point cloud map without dynamic objects in real time.In the experimental scene of 90m×90m,the dynamic environment can be described by updating the hierarchical cost map in real time.(3)In order to reduce the deviation between the actual trajectory of the mobile robot and the global path,this thesis proposes a bidirectional A~* algorithm based on angle constraints and a DWA algorithm based on adaptive prediction time.On the basis of the A~* algorithm,the steering angle constraints,the angle constraints of the starting point and the target point are added,the two-way search mechanism is combined to improve the efficiency of the algorithm’s path finding.In the trajectory evaluation link of the DWA algorithm,the adaptive prediction time is used to generate the predicted trajectory.In the MATLAB simulation experiment,the algorithm in this thesis can quickly find the path that meets the constraints,and the average deviation distance of the algorithm in this thesis is 16% of that of the traditional algorithm.In the physical experiment,the global path planning algorithm in a map with 450×450 nodes,the average total planning time is about 100 ms,which meets the requirements of real-time planning.(4)In order to improve the safety of mobile robots when avoiding dynamic objects,this thesis combines the prediction results of dynamic obstacles with local planning,and realizes autonomous navigation of mobile robots in dynamic environments through global path replanning.Through simulation experiments and experiments on the experimental platform,it is proved that the path planning algorithm in this thesis has better and safer obstacle avoidance effect when facing dynamic obstacles.In order to verify the effectiveness of the method in this thesis,this thesis builds an experimental platform on a four-wheel drive chassis,and deploys the above algorithm in an industrial computer.On the experimental platform,lidar is used to realize various functions of autonomous navigation of mobile robot in unknown dynamic environment,and the effectiveness of the algorithm in this thesis is verified by experiments.