A Simple yet Efficient Dynamic System for Robot Path Planning

Abstract

In this paper, a simple yet efficient dynamic system is proposed for real-time collision-free robot path planning. Like most robot path-planning approaches, the environment is represented by a topologically organized map. Each grid point on the map has only local connections to its neighboring grid points. Thus the computational complexity linearly depends on the number of grid points. Unlike the neural network based path-planning model proposed by Yang and Meng (2001), the proposed algorithm efficiently propagate the distance from the target to the entire robot workspace. Thus the real-time optimal robot path is generated through the dynamic distance landscape without explicitly searching over the global free workspace nor the collision paths, without explicitly optimizing any global cost functions, and without any prior knowledge of the dynamic environment. Therefore the model algorithm is computationally efficient. The effectiveness and efficiency of the proposed dynamic system are demonstrated through simulation and comparison studies