Research On Autonomous Exploration And Motion Planning Methods Of Wheeled Mobile Robot

At present,wheeled mobile robots are developing in the direction of autonomy and intelligence.Wheeled mobile robots with autonomous sensing,movement,and decision-making capabilities have a broad development space and have become important equipment for manufacturing and logistics.Autonomous navigation technology is a crucial technology to realize the autonomy and intelligence of wheeled mobile robots,while autonomous exploration and motion planning are essential components of autonomous navigation technology.Autonomous exploration enables wheeled mobile robots to actively perceive and build environment maps,while motion planning,as a bridge between perception and control,can promote the high-quality motion of wheeled mobile robots.The research of autonomous exploration and motion planning methods is highly significant to enhance the autonomy and intelligence of wheeled mobile robots,and is the key to further promoting wheeled mobile robots into more practical application scenarios.However,the current research on autonomous exploration methods is characterized by a single pathway for frontier region detection,lack of inefficient frontier region rejection mechanism,repeated path planning,and underutilization of environmental semantic information,which restrict the further improvement of autonomous exploration efficiency.In the current motion planning method,the low efficiency of planning in narrow regions,the lack of trajectory safety and smoothness,the lack of consideration of robot dynamics in geometric path planning,and the need for multiple adjustments to get the dynamics-constrained trajectory limit the high quality of robot motion.To address the above issues,this thesis carries out algorithm design,system development,and validation related to autonomous exploration and motion planning with a wheeled mobile robot as the research object.The main work of the paper includes:(1)To solve the problems of inefficient autonomous exploration due to a single pathway for frontier region detection,lack of inefficient frontier region rejection mechanism,and repeated path planning,an efficient autonomous exploration method based on hybrid frontier point detection and multi-path evaluation is proposed.First,the hybrid frontier detection and maintenance method is designed to fully acquire the frontier regions in the environment by analyzing the Lidar data and processing the data-changing regions on the occupancy grid map.To maintain effective frontier regions,an inefficient frontier point rejection strategy is designed using the local update mechanism of the Euclidean distance map.Second,to avoid repetitive path planning and accurately obtain the path cost,a novel multi-path planning method is proposed to generate multiple paths from the robot to the unexplored frontier region that are safe and favorable for observation by the robot’s self-loaded sensors.Then,a multi-objective utility function is designed to evaluate the optimal path and use an iterative optimization strategy to improve its smoothness.Finally,this smoothed path is followed using a nonlinear model predictive control method.The experimental results show that the method in this chapter reduces the exploration path length by 27.07%and the exploration time by 27.09%on average compared to the baseline methods.(2)To facilitate wheeled mobile robots to efficiently complete exploration tasks using environmental semantic information,a state-adaptive semantic autonomous exploration method for mobile robots in general indoor environments is proposed.Firstly,a frontier point simplification and maintenance method considering map-building features is proposed to quickly extract frontier points and lay the foundation for subsequent decision-making sessions.Combined with the door detection and correction module and the scene semantic information recognition module,the state update and maintenance method is designed for the robot exploration process.Then,the adaptive exploration method based on robot state is proposed triggering different exploration strategies based on different robot states to achieve adaptive adjustment of exploration strategies according to the exploration needs in different scenes.Second,the proposed path optimization method can obtain a path that promotes the safe and smooth operation of the robot,while the path retains the beneficial characteristics of the original path.The experimental results show that the proposed strategy can accelerate the exploration activities of the robot in the general indoor environment and significantly reduce the behavior of repetitive exploration.The exploration path length and exploration time are reduced by 15.6%and 10.8%,respectively,compared to the integrated metrics approach.The exploration time is reduced by 18.9%compared to the semantic exploration method using heuristic objects.(3)To realize safe and smooth navigation operations for wheeled mobile robots,a safe and smooth motion planning method based on a safe connect rapidly-exploring random tree with cubic spline interpolation is proposed.The method consists of two parts:the safe connect rapidly-exploring random tree(SC-RRT)path planning method and the trajectory planning method based on the time allocation mechanism and the cubic spline interpolation method.Firstly,the SC-RRT method is proposed to find a path with a good safe distance from obstacles quickly by branch growth toward the environment safety area and bi-directional interconnection mechanism.Second,the design of critical path point selection and the initial path point time allocation strategy based on the S-shaped acceleration and deceleration method are used to complete the initial time allocation of critical path points.Then,the smooth trajectory with time as the parameter is obtained by using the cubic spline interpolation method.Finally,after further processing by dynamics feasibility detection and time reallocation mechanism,a collision-free and smooth trajectory that can be executed by the robot can be generated to realize safe and smooth robot navigation.(4)To perform simultaneous path and trajectory planning and directly generate trajectories satisfying dynamical constraints,a dynamically feasible B-sample random tree(DB-RRT)method combining the B-sample convex hull property and the fast expansion capability of the Rapidly-exploring random tree method is proposed.The method can solve the path and trajectory planning problem simultaneously and directly generate the trajectory without collision and satisfying the dynamical constraints.The sustainable growth capability of the DB-RRT tree is analyzed,and then the concept of a dynamically feasible region is proposed.And using the dynamically feasible region to further improve the planning efficiency,a geometric method is designed to determine whether a dynamically feasible trajectory is generated or not,and a steering function is designed to guide the growth of the tree.The fast marching guided dynamically feasible B-sample random tree(FMDB-RRT)method can further improve the efficiency of trajectory generation and enhance the safety of trajectories.In the building environment,the planning time is reduced by 22.9%and the average clearance to obstacles is improved by 46.0%compared to the path-guided approach.The application testing of the autonomous navigation wheeled mobile robot developed by relying on the autonomous navigation system and autonomous exploration and motion planning methods in this thesis in a factory environment is introduced.