Dynamic path planning under randomly distributed obstacle environment

摘要

To achieve an online robotic path planning system which can dynamically react to environmental changes, the study proposes a real-time system which updates locally the input information and re-plans in time an alternative path for consecutive following under restrained computing capability of the agent. The system consists of a framework of functions to switch the working space successively into so-called feature space and topological space, and then pave the way from a initial position I to an objective position O. Based on discretization of the obstacles in the working space, the study first abstracts the configuration obstacles as a set of randomly scattered points, and forms the feature map. Merging together the elementary transformation between Delaunay triangulation and Voronoi diagram, the scattered points can be connected to tessellate a topological map. By employing further the generalized Voronoi diagram and D* Lite path planner, the developed system can thus generate a shortest path for obstacle avoidance in the changing environment. With a sequence of simulations, the developed system showed a promising capability of path replanning which admitted for future applications.