Research On Autonomous Navigation System Of Unmanned Vehicle In Indoor And Outdoor Environment

Unmanned vehicles have become the research hotspot of mobile robots,and are playing an increasingly important role in intelligent inspection,agricultural production,and safe travel.Autonomous navigation is one of the key technologies to achieve autonomous and intelligent unmanned vehicles.However,at present,autonomous navigation of unmanned vehicles is mostly conducted in a single indoor or outdoor environment,which cannot be applied to different indoor and outdoor environments.This thesis designs and independently builds an autonomous navigation system for unmanned vehicles in different indoor and outdoor environments through research on three aspects of unmanned vehicle localization and mapping,freespace detection and layered costmaps,and path planning,effectively solving the autonomous navigation problem of unmanned vehicles in different indoor and outdoor environments.The main research of this thesis is as follows:(1)Aiming at the problems of accumulated errors in outdoor long-distance mapping and inaccurate positioning due to the absence of GPS indoors,this thesis proposes a localization and mapping method based on multi-sensor fusion.Firstly,Li DAR and IMU are jointly calibrated to achieve the unification of the coordinate system;Then,based on the LIO-SAM algorithm,a loop closure detection method using Scan Context spatial descriptors is added,and a multi-factor graph model optimized by GTSAM is used to complete the task of constructing and locating 3D point cloud maps in indoor and outdoor environments;Finally,three laps of data were collected and the tracks of the front and rear laps were analyzed using EVO tools,resulting in a reduction in absolute pose error of 50.59% and a reduction in relative pose error of 94.34%.At the same time,comparing the tracks generated by the improved algorithm before and after,it was shown that this method can reduce the outdoor accumulated error and inaccurate indoor positioning issues.(2)Aiming at the problems of inability to construct freespace and single semantic information in indoor and outdoor 3D point cloud maps,this thesis proposes an indoor and outdoor freespace detection and layered costmaps update and creation method.Firstly,the 3D point cloud map is divided into concentric regions,region plane fitting,and ground likelihood estimation to complete the ground point cloud segmentation;Then,use the Cloud Compare tool to remove the dynamic noise from the point cloud map after segmentation of the ground,and use radius filtering and pass-through filtering to project the processed 3D point cloud map into a two-dimensional grid map,completing the construction of a freespace;Finally,the main costmaps is generated by combining the static layer,obstacle layer,and expansion layer to solve the problem of single semantic information in the single layer map during the navigation process.(3)Aiming at the problems of large memory resource consumption and large path curvature in outdoor large scenes and indoor long corridor navigation path search,this thesis proposes a dynamic weighted smooth JPS global path algorithm and a local path planning based DWA algorithm.Firstly,based on the traditional JPS algorithm,a dynamic weighted heuristic function is added and a third order Bezier curve smoothing process is performed to accelerate path search and reduce memory consumption.DWA algorithm is used to achieve static and dynamic obstacle avoidance indoors and outdoors;Then using MATLAB to simulate and analyze the improved algorithm,the results show that the improved algorithm has less search time,and the planned path is smoother;Finally,the improved algorithm is transplanted to the Move＿Base navigation framework for real vehicle experiment.The results show that unmanned vehicles can achieve dynamic obstacle avoidance in indoor corridors,floor navigation,narrow space navigation,outdoor long-distance navigation,and indoor to outdoor autonomous navigation.