ABF-RRT*: Bidirectional fast optimal rapidly-exploring random tree via adaptive sampling of a permissible hyper-ellipsoid

This paper proposes a global path planning algorithm based on adaptive sampling and bidirectional fast exploration, ABF-RRT*. ABF-RRT* introduces an adaptive sampler that dynamically constructs a hyper-ellipsoid according to the progress of dual-tree exploration and adaptively weighs the sampling inside and outside the specified area, which effectively guides the process of dual trees advancing each other. The improved bidirectional planner combines the bisection optimization process, significantly improves the topological structure of the random tree, and accelerates the convergence to the optimal path solution. Since the algorithm relies on the tree expansion framework, it is compatible with various samplers and graph pruning strategies. Through extensive simulation experiments, the superiority of the two cores is analyzed separately, and it is proved that ABF-RRT* has excellent performance and efficiency in path planning in various environments.