Increasing Diversity of Solutions in Sampling-based Path Planning
Vojtěch Vonásek, Martin Saska · 2018
Sampling-based path planning algorithms like Rapidly-exploring Random Trees (RRT) are widely used to solve path planning problems for robots with many Degrees Of Freedom (DOF). Although the sampling-based planners are stochastic and provide different solutions in every trial, many resulting trajectories can be quite similar, e.g., because they are close to each other. Computation of diverse paths, i.e., paths leading through different parts of the configuration space, is important for online navigation of mobile robots or for finding solutions leading through different narrow passages of the configuration space. In this paper, we propose an extension of the RRT algorithm to find diverse paths in the configuration space iteratively. The idea of the proposed method is to prohibit the exploration of selected regions of the configuration space. The prohibited regions are defined using the waypoints of paths computed in the previous iterations. We show that the proposed method finds more diverse trajectories than can be achieved by repeated computations of a single sampling-based planner.