Research On Mapping Of Service Robot In Dynamic Environment

More and more service robots are gradually integrating into people's lives.While bringing convenience to people and helping alleviate social pressures such as providing for the elderly and helping the disabled,they are also facing new challenges.One of them is how to deal with dynamic and quasi-dynamic obstacles,such as fast-moving people,open and closed doors,in the environment while mapping.The existing mapping algorithm,that is,simultaneous localization and mapping(SLAM),is proposed on the assumption that the environment is static.However,there is less research on the map construction method in the environment that contains dynamic and quasi-dynamic obstacles,that is,the dynamic environment.Therefore,this article is oriented to service robots and studies its mapping methods in dynamic environments.This paper combines the existing two-dimensional LIDAR-based SLAM,uses a grid map to represent the real environment and graph optimization framework,which is divided into two parts: the front end and the back end.The front end will preprocess the laser point cloud data and obtain the pose of the robot through the scan match algorithm.And the back end uses the least square method to optimize the robot's pose by constructing pose graph.This paper proposes corresponding improvement methods on laser point cloud data preprocessing,data scanning matching and map building model,respectively,and adopts the loop detection method based on local map.Under the graph optimization framework of SLAM,mapping in dynamic environment is realized.The main work of the dissertation is as follows:(1)In view of the noise problem of laser observation data under dynamic environment,an improved denoising algorithm based on adaptive breakpoint detection is proposed.The algorithm calculates the distance of adjacent laser points to the origin,respectively,and predicts the distance from the next laser point to the origin,then according to the distance difference between the predicted distance and the real distance,the laser point cloud data is quickly divided into different data fragments,which with fewer data points are removed to complete the denoising and avoid the influence of noise on mapping.(2)Research on the map-based laser scanning matching method,and propose an improved correlative scanning matching algorithm.The algorithm first sets the search window and search step on the grid map,and gradually searching to find the optimal pose,and then performs Lanczos interpolation on the grid map to make the grid map continuous and derivable.Finally,the laser point cloud data is converted to the grid map coordinate system and the least square term is constructed to optimize the robot pose to be more accurate.(3)In the fact that the slow-moving obstacles may exist in the environment,an improved counting model is proposed to construct local and global grid maps and eliminate the quasidynamic obstacle trajectory in the created map.The local map will be saved for use in the loop detection process.Different from the way of searching the data link,the loop detection in this paper includes two steps: the first step is to find the local maps closest to the current pose,and the second step is to use an improved scan matching algorithm to detect whether there is a loop.In order to speed up the detection speed,a branch and bound algorithm is used to quickly search in the search window to find the most suitable pose.And the loop closure detected is added as constraints to the back end,then all poses are optimized,and the map is updated.The algorithm proposed in this paper is tested and verified in a dynamic environment.(4)In actual scenarios,the robot experiment platform is used for experimental verification.First,it is verified that the denoising algorithm proposed in this paper can effectively remove the noise points in the laser point cloud data.Then,in the presence of dynamic obstacles,the robot's pose is positioned through laser scanning matching,and an environment map is constructed to verify the robustness and effectiveness of the improved scanning matching algorithm proposed in this paper.Finally,in the presence of dynamic and quasi-dynamic obstacles,an environmental map is constructed,and the trajectory of quasi-dynamic obstacles is effectively eliminated at the same time,verifying the effectiveness and feasibility of the improved counting model proposed in this paper in eliminating the trajectory of quasi-dynamic obstacles.The above experiments also proved the effectiveness and adaptability of SLAM proposed in this paper.