Physical potential adaptation for real-time robot path planning

Abstract

Artificial Potential Field (APF) is a method widely used for an autonomous robot path planning and navigation because of its light complexity and elegance of results. Especially in low power with real-time required path planning, APF is useful method such as swarm robot, robot soccer, or mobile sensor node. APF defines potential throughout the field to attract robot to the goal while repulsion robot from obstacles. However, it shows limitation to applying to real driving with environment reflecting to their path planning and multi-robot support in distributed path planning because APF defines only attract force and repulsive force. For this problem, there is no work for adapting APF to environment with constrained condition. In addition to this, existing APFs which can apply to multi-robot, firstly, just regard robots as obstacles even if these robots are not obstacles, secondly, focus on special circumstance, not considering environmental reflecting or not consider characteristic of object in path planning. To solve this limitation and extend, this thesis proposes a physical potential adaptation APF by defining new potential factors for adapting and extending APF whose idea is inspired from physical formulae. In addition to this, I try to solve local minimum problem which exist potential minimum locally even if it is not global minimum, which are problems of limitations from APF. Physical potential adaptation APF includes environment sensitive potential factor, multi-robot planning potential factor, and local minimum escaping factor. Therefore through physical potential adaptation APF, this paper enables distributed and environment sensitive reflecting path planning in multi-robot with real-time using the suggested APF. All of these works are verified its effectiveness with the simulations by Player/Stage simulator.