Active Mapping And Obstacle Avoidance In Navigation For Intelligent Vehicle Based On Visual Localization

Simultaneous localization and mapping(SLAM)algorithm is widely used in the field of intelligent robots,so that intelligent robots can locate themselves and build maps in unknown environments.SLAM algorithm can be divided into laser slam and visual slam according to different sensors.Visual slam has the advantages of low hardware cost and rich search information dimensions.It has unique technical advantages and great potential in the field of slam.However,the traditional SLAM algorithm needs the assistance of operators,and there are technical shortcomings in unmanned inspection,disaster rescue and so on.Therefore,according to rgb-d vision and autonomous exploration strategy,this paper carries out the research on active mapping and navigation obstacle avoidance of intelligent vehicle based on visual positioning technology.Specifically include:(1)A position and attitude estimation method of visual inertial odometer based on point and line features is studied.In this paper,the feature extraction accuracy of slam will be better than that of the existing camera,which will lead to the problem of losing the matching accuracy between slam and the existing feature extraction system.(2)An active exploration strategy based on boundary guided closed-loop is studied.Aiming at the problem that closed-loop optimization in traditional passive mapping needs the help of operators,an active visual slam is designed,which relies on the known and unknown boundaries of the map as a guide to explore and form closed-loop optimization in an unknown environment.(3)A navigation obstacle avoidance method based on gravity constraint is studied.Aiming at the problems of excessive body angle and difficult decision-making when the robot passes through obstacles in the traditional a * path planning algorithm,an improved algorithm based on gravity constraint control is designed.In the process of path planning,the "repulsion force" of obstacles is considered to realize the active and safe collision avoidance of intelligent car.At the same time,it is combined with local path planning to avoid gravity traps.(4)The hardware platform and software framework are built,the simulation verification is carried out on the ROS robot development platform,and the designed control strategy is verified on the spot through the built hardware platform.