A reactive planner for multiple robots involved in competitive tasks

In this article, we propose the concept of dynamic map (DM), which allows to represent the intrinsic kinetic capabilities of a vehicle, and therefore to control its displacements once the geometry of this map is known. This concept is based upon the theory of reachable regions for a dynamic system. The work presented in this paper concerns the aspects of: (1) the construction and representation of the map, (2) the learning by a robot of its own map, and (3) the description of the sensori-motor relation between the visual information received by the robot and its displacement using the DM. The second part of the paper concerns the exploitation by a robot of the information contained in its own DM and that of other robots. The maps are combined in order to define a new distribution of the robot´s reachable space, in the context of the task at hand. An optimization algorithm can then be used to find a path that takes into account all robots capabilities. This planning scheme is applied to the case of a single robot that is pursued by one or several other robots and therefore must plan an escape trajectory