Path Planning Methods For Nonholonomic Constrained Systems Based On Rapidly Exploring Random Trees

In the trend of increasing application scenarios of mobile robots,the demand of intelligent robots continues to increase.Since path planning algorithms can affect the safety and execution efficiency of robots during autonomous motion,it is still a research focus about performance optimization of path planning algorithms in academia and industry.Rapidly Exploring Random Trees(RRT)path planning algorithm has become one of the most popular sampling-based path planning algorithms because of its probabilistic completeness,simple structure,and strong searching capability.The working scene of mobile robot is analyzed in this paper,and the wheeled mobile robot satisfying nonholonomic constraints is seen as the main research object.We analyze some the path planning problems in structured static obstacle environment,dynamic obstacle environment and barrier-free two-dimensional manifold S~2using RRT algorithm.Three research contents are as follows.(1)A path planning method called CRAB-RRT(Circle Restraint Adaptive Bias RRT)is proposed for offline path searching in structured static obstacle environments with different obstacle densities.The method reduces the number of samples in invalid areas and minimizes the generation of ineffective nodes by using a circle-biased sampling strategy.Selecting the appropriate nearest point using Octile distance and angle constraint function,and it can solve the problem of large computational time when using Euclidean distance.Target attraction component and adaptive extension step length are added to promote tree node biased extension.Using constrained path smoothing strategy to eliminate redundant turn points on the initial path and smooth the path.Numerical experiments verify the feasibility and rationality of the CRAB-RRT algorithm,and simulation results show that the performance of the CRAB-RRT algorithm is improved to some extent.The algorithm is applied to an experimental vehicle that satisfies nonholonomic constraints,and the experimental vehicle can safely and effectively follow the path and move to the target area.(2)A path planning method called DIB-RRT*(Dynamic Informed Bias RRT*)is proposed for searching safe paths in dynamic obstacle environments.Although the RRT*algorithm has good flexibility,dynamic performance,and considers asymptotic optimality conditions,it still faces inefficiency in planning initial path,large sampling space,many invalid sample points and long path searching time.The dynamic connected domain sampling strategy is used to optimize the search efficiency and path length of the initial path.The adaptive extension of new nodes is realized by combining the artificial potential field method with adaptive extension step length.A path repair method based on obstacle information is used to reduce repair sampling time and achieve local path repair.Numerical experiments are conducted to compare and verify the effectiveness and feasibility of the DIB-RRT*algorithm,and simulation results show that the DIB-RRT*algorithm achieves rapid local path repair and dynamic obstacle avoidance function.(3)A path planning method called SBF-RRT(Spherical Barrier-Free RRT)is proposed for searching feasible paths on the two-dimensional manifold S~2.The method uses parameterization and inverse transform sampling methods to solve the problem of uniform random sampling on the S~2manifold.The approximation distance metric function of the S~2manifold surface is obtained through angle cutting and piecewise summation methods.New nodes on the S~2manifold surface are generated using the angle cutting and coordinate projection methods.The directed connection problem between two points on the S~2manifold is solved using continuous interpolation methods.Numerical experiments are conducted to validate the effectiveness of the SBF-RRT algorithm.