Heat diffusion method for intelligent robotic path planning

Abstract

Real-time collision-free motion planning is an important issue in many autonomous systems including robotics and intelligent systems. It endows intelligent robotic systems with an ability to plan motions and to navigate autonomously. This ability becomes critical particularly for robots which operate in dynamic environments, where unpredictable and sudden changes may occur. This paper presents an intelligent method for real-time robotic path planning by using potential field data generated by simulating heat diffusion. Heat is conducted throughout the field from the objective location, and produces a gradient (potential field) which is used for path planning. By iteratively following the highest temperature gradient, the optimal path between a robot and its objective can be established. Both steady-state and transient heat diffusion are studied. By utilizing the iterative process of transient heat diffusion during potential field generation, path planning in a dynamic environment becomes more feasible than steady-state diffusion. A computer program is built using Java code to simulate dynamic obstacles/environments and to generate potential fields upon which to enable dynamic path planning. Techniques to enable obstacle avoidance (for obstacles lying linearly between the robot and its objective) are examined. The introduction of obstacles that allow heat to diffuse through them (however do not allow robot passage) and the introduction of multiple heat sources to a potential field produce a problem known as 'stalling' at the 'local maximum'. A basic iterative solution of replicating fake obstacles is developed and described that may be expanded upon to effectively solve the problem of stalling at the local maximum, hence enabling dynamic robotic path planning to be successfully performed using a simple and quick transient heat diffusion algorithm.