| With the continuous development of sensor technology,unmanned autonomous robot can replace human to complete tasks such as work and exploration in more and more environments.A single sensor has been difficult to meet the needs of mobile robot to complete positioning and navigation in the complex working environment.Based on the analysis of the research status of SLAM algorithm at home and abroad,this paper optimizes the self positioning and map creation ability of mobile robot by integrating lidar and binocular camera,and builds a set of mobile robot positioning and navigation system.The specific research contents of this paper are as follows:(1)A mobile robot platform equipped with two-dimensional lidar and binocular camera is built.The mobile robot motion model,wheel odometer,binocular camera ranging model and lidar measurement model are mathematically modeled,and the wheel odometer and binocular camera are calibrated.According to the requirements of the robot,the information processing system is designed.The lower machine system is responsible for the motion control of the mobile robot platform and the reception of the built-in sensor data.The upper computer system receives the data of radar sensor and visual sensor,and runs the upper algorithm.(2)Karto SLAM algorithm is improved for 2D lidar mapping,and a point cloud preprocessing algorithm based on hybrid filtering is proposed to correct invalid points,outliers and error points caused by equipment accuracy and environmental factors.Firstly,the pass through filter is used to eliminate the invalid points,then the radius filter is used to identify the outliers and error points,and finally the Kalman filter is used to correct the value of the error points.A series of experiments are carried out to test the role of important parameters in the point cloud preprocessing algorithm,and the optimal parameters are selected.Finally,the KARTO SLAM algorithm before and after the improvement is tested in the field.The improved algorithm can effectively reduce the ghosting phenomenon in map construction.(3)In order to integrate laser mapping and visual mapping,the architecture of orb-slam2 system is studied,and the principles of extraction and matching of orb feature points,pose estimation,local map tracking and loop detection are analyzed.In the original algorithm framework,the octree map construction part and grid map construction part are added to solve the problem that the sparse point cloud map constructed by ORB-SLAM is difficult to be used in path planning.The cumulative fraction method The integration method is used to fuse the grid map constructed by lidar and the grid map constructed by binocular camera for path planning later.(4)Aiming at the problem that PRM path planning algorithm runs slowly and is difficult to sample narrow channels,a grid probability path graph algorithm is proposed in this paper.Firstly,the grid is used to divide the map,and according to the area of obstacles in the grid,the grid threat level is divided,and different sampling strategies are used accordingly.Secondly,a resampling method that the sampling point falls in the obstacle is proposed,which improves the sampling efficiency and increases the sampling in the narrow channel.At the same time,the connection strategy is changed.When connecting sampling points,it no longer traverses all points,but only connects with nearby grids,which reduces the time-consuming of the algorithm.After generating the path,the path is optimized and smoothed to improve the path quality and make it conform to the motion constraints of the mobile robot.Through simulation analysis,the basis for the selection of grid scaling coefficient K in grid probability path graph method is obtained.The simulation results show that the operation time and success rate of the grid probability path graph method are improved compared with the classical algorithm.The grid probability path graph algorithm is deployed on the mobile robot platform to complete the field path planning. |
