Autonomous Environment Exploration Of Mobile Robot Based On Grid-Octree Hybrid Map

Autonomous mobile robot is an important research area of robotics.It is an essential capability for autonomous mobile robot to explore the environment through sensor information in an unknown environment and build corresponding maps at the same time.Aiming at the problem of autonomous environment exploration and map construction of mobile robots in unknown indoor scenes,this thesis mainly researches from three aspects,including simultaneous localization and mapping,autonomous exploration strategies and path planning.Simulation verification and mobile robot experiments are also performed.The main research content completed are as follows:(1)Simultaneous localization and mapping part is responsible for obtaining the pose information of the mobile robot and the grid-octree hybrid map.This thesis uses the improved Rao-Blackwellized particle filter algorithm to estimate the pose of a mobile robot.In order to achieve rapid navigation of the robot and obtain richer environmental information,a grid-octree hybrid map of the environment is created simultaneously using laser and depth camera information.(2)The autonomous exploration strategy part is responsible for selecting the target pose for exploration.This thesis proposes an autonomous exploration strategy based on a hybrid candidate extraction algorithm,including the extraction and grouping of frontiers,the extraction of candidates,and the evaluation of candidates.The frontiers are extracted and grouped by octo-map,and the central frontier of each group are used as the key frontiers.Aiming at the limitation of the single candidate extraction algorithm,this thesis proposes a local candidate extraction algorithm based on geometric rules and a global candidate extraction algorithm based on accessible space.Combining these two algorithms,a hybrid candidate extraction algorithm is proposed,which can take into account both the speed of autonomous exploration and the traversal of the environment.Comprehensively considering driving distance,rotation angle,and information gain,this thesis constructs a benefit function to evaluate candidates.For local candidates and global candidates,the cyclic traversal algorithm and the Bayesian optimization algorithm are used to select the optimal candidate as the target.(3)Path planning part is responsible for driving the robot to the target pose.Aiming at the complex and changeable indoor environment,this thesis uses the hierarchical costmaps to model the environment,and uses two-dimensional laser information and three-dimensional point cloud information for obstacle detection.This thesis uses A*algorithm for global path planning,improves dynamic window method based on forward motion constraints,and uses improved dynamic window method for local path planning.Based on the ROS(Robot Operating System)function package,the robot's autonomous navigation is realized,so that the robot can safely and efficiently travel to the target in an indoor environment.(4)Finally,the performance of the autonomous environment exploration system in this thesis is simulated in the Gazebo simulation environment.A Turtlebot2 robot equipped with a LRF and a depth camera is used as the hardware platform,and ROS is used as the software framework to build an experimental platform.Further experiments verify the effectiveness of the autonomous environment exploration system in this thesis.