Unmanned vehicle path planning is a prominent field in the current era, giving rise to numerous related algorithms. One such algorithm is the RRT* algorithm, which is widely recognized. However, this algorithm involves random sampling, resulting in an excessive number of nodes and paths with numerous inflection points that are not smooth enough for application in unmanned vehicle driving. This paper aims to address these issues and propose improvements. Aiming to enhance the traditional RRT algorithm in path planning, although it can improve search efficiency, it will increase the number of expansion nodes within obstacle areas. To address this issue, it is suggested to incorporate a pruning algorithm during the generation of sampling nodes. The principle of the pruning algorithm is to reduce the number of nodes in the search tree by eliminating the "unimportant" search nodes during the search process, while striving to maintain the accuracy of the search tree. For the issue of the planned path not being smooth and containing too many inflection points, a trajectory optimization method based on the Floyd algorithm is proposed. This method aims to smooth the generated trajectory to ensure a more continuous curvature. Simulation experimental results demonstrate that incorporating the pruning algorithm reduces a significant number of generated nodes in the search tree, enhances the speed of path planning through new search rules, and results in a curve after optimization that is more suitable for unmanned vehicles to navigate. This optimized curve is superior to the original curve in terms of both length and smoothness. Finally, the RRT* fusion algorithm, which combines the Floyd algorithm and pruning algorithm, is implemented on the MATLAB platform for experiments. In the same planning task, the new fusion algorithm reduces the path planning time of the original algorithm by 52.37% and decreases the number of inflection nodes by 75%. This improvement can meet the increased real-time path planning demands of unmanned vehicles during task execution.
|