Research on slam and path planning based on ros robot

Abstract

With the development of artificial intelligence technology, human-computer interaction has become the trend of robot research, navigation robots are widely used. Navigation robots autonomous navigation in an unknown environment, including perception of the environment, autonomous positioning and dynamic obstacle avoidance, etc. Among them, Simultaneous Localization and Mapping (SLAM) and path planning are prerequisites for achieving robot intelligent navigation. Aiming at problems such as low positioning accuracy and inability to accurately sense obstacles in 3D complex environments when using lidar. Firstly, based on the data characteristics of lidar and depth camera, the projection of point cloud collected by depth camera on 2D plane and lidar point cloud are fused by Bayes formula, then build an occupied grid map of the indoor complex environment. In addition, the traditional artificial potential field (APF) is easy to fall into the local optimal solution and oscillation of the potential field, and there may be collisions in the presence of dynamic obstacles. In the face of complex dynamic environments, it is difficult to plan a complete path. The velocity factor of obstacles is added to judge the movement trend of obstacles in the APF method, and a new repulsion function is introduced to overcome the local optimal trap and avoid the problem that obstacles near the target point lead to the target unreachable. Experiments show that the fusion raster map has higher precision than the single laser radar in the establishment of grid maps. In the aspect of path planning, the improved APF method can effectively solve the stochastic dynamic obstacles, and improve the navigation efficiency of robots.