Research On Monocular Vision Aided Localization Algorithm And Path Planning

In recent years,the rapid development of sensors and artificial intelligence has provided an opportunity for the expansion of robot functions.Autonomous mobile robots have become an indispensable part of human life and production.Realize map construction,positioning and navigation module robot path planning function function of the other top,its stability and robustness is important basis of evaluation of the intelligent robot and work performance,so how to improve the positioning precision of the navigation module,positioning stability and environmental adaptability has been a hot issue in the field of robot.In this paper,the navigation function of mobile robot in the environment with similar features is taken as the research background.In view of the limitations of the adaptive Monte Carlo(AMCL)positioning algorithm,the real-time performance of path planning and the accuracy of map building,the function of the navigation module is studied.The main research contents of this topic are as follows:(1)An improved AMCL algorithm that can correct the positioning results in real time is proposed,which solves the problem of large positioning error caused by inaccurate odometer information and improves the global positioning ability of the algorithm.In this paper,monocular vision is used to obtain label information in the environment and extract visual features online.Machine learning method is used to obtain absolute position information by text classification.The positioning results of AMCL algorithm are corrected in real time to improve the positioning accuracy and robustness of the navigation module.(2)In view of the real-time performance of path planning and the rationality of trajectory,the characteristics of A*algorithm with different cost calculation formulas were studied by combining simulation and experiment.The method of combining global path planning and local path planning is adopted to solve the driving trajectory in real time,and a more real-time path planning algorithm is configured for the navigation module based on the actual working environment.Based on different sensor configuration schemes,the map construction effects of different SLAM algorithms in the actual environment are studied to provide a basis for sensor configuration and SLAM algorithm selection when the robot is mapping in an unknown working environment,and to improve the environmental adaptability and mapping accuracy of the navigation module.(3)To build an experimental platform that can realize autonomous movement,including hardware circuit and software control.The chassis is driven by two wheels with differential speed,and the chassis movement is controlled manually by handle control and wireless remote control.The sensing module that can access multi-line radar,single-line radar,monocular vision,IMU and encoder is configured,and the power management module is designed for the complex electrical system to ensure the stability of the experimental platform when working.The software system uses ROS robot operating system to receive real-time data operation from the sensor and data transmission efficiency of the sensor through the communication mechanism of distributed management.(4)The autonomous moving experiment is carried out on the experimental platform,which verifies that the AMCL algorithm with monocular vision is improved in positioning accuracy and global positioning ability.A variety of global path planning algorithms were tested with paths of different lengths to verify the real-time performance of the improved path planning algorithm.The safety of the navigation system was verified by dynamic obstacle avoidance experiments.