Dynamic Path Planning for Rapidly Expanding Autonomous Vehicles in Complex Environments

To address the problems of slow convergence and low path smoothness of path planning algorithms in complex environments, we propose a two-layer path planner consisting of a global layer and a local layer. In the global layer, a precise sampling rapidly-exploring random tree (PSRRT $^{\ast }$ ) algorithm is proposed to address the problem of slow convergence speed of path planning algorithms. First, a precise sampling model is established to obtain precise path points. The model enhances the directional guidance of the sampling points and reduces the number of invalid sampling points. Second, an adaptive path node expansion model is established to obtain the initial global path. The expansion model constructs a new gravitational field based on the precise sampling points and adaptively adjusts the repulsion field to quickly obtain an initial global path. Finally, the initial global path is pruned and smoothed to obtain the optimal global path. In the local layer, a goal-directed dynamic window approach (GDWA) is designed to enhance path smoothness. The GDWA uses a novel goal-oriented evaluation function to produce the optimal local path. The experimental data and simulation results show that the path planner designed in this article can greatly reduce the planning time and computation, and improve the path smoothness in complex environments. A real environment is set up to verify the feasibility of the algorithm.