Mobile Robot Navigation In A Dynamic World Using An Unsteady Diffusion Equation Strategy

This paper discusses a method for map and sensorbased planning and execution, i.e. navigation, of smooth collision-free robot motion paths. Application of an unsteady diffusion equation model to path planning in a time-varying world distinguishes this approach from prior work in the field. Collision-free robot paths between a start and a mission dependent goal point are generated very rapidly by on-line simulation of a diffusion process and by repetitively evaluating the gradient of the computed concentration distribution functions. Related numerical algorithms are appropriate for implementation on a massively parallel distributed processor architecture. This paper describes application of the approach to various scenarios of robot vehicle navigation. Results of a field experiment performed with the mobile robot MACROBE are presented.