Reactive deformation of path for navigation among dynamic obstacles

Abstract

This paper presents experimental results of a reactive deforming path algorithm, based on elastic strips, in a 2D/3D simulator. This algorithm generates motion plans that respect global constraints while making sure that the robot avoids collisions with all stationary, moving and unexpected obstacles. The problem is addressed by integrating a global planning algorithm with a local reactive approach. The planner performs path planning in two stages: (i) Global planning of the path is done using a heuristic search algorithm such as A*. (ii) In a dynamic environment, with several obstacles moving around in an unpredicted fashion, the planner deforms the path reactively by subjecting it to artificial forces. For making sure that the robot is always on a collision free path, the plan is incrementally updated with changes in the environment. The plan representation uses deformable links connecting the robot's configurations along the path, stretching or retracting the path as required to avoid collisions.