Avoiding collision logjams through cooperation and conflict propagation

Abstract

A methodology for resolving collision conflicts of many fast moving robots involving distributed cooperative behavior is proposed. Robots are unaware of the goals and plans of other robots. However, each robot is assumed to know the current states of other robots within a given range of vision. The current state of a robot is represented by its current and target velocities and its heading direction. A conflict issue between two robots is a detection of an upcoming collision. The resolution of the conflicts occurs at three levels. At the first level, the robots attempt to resolve the conflict individually. If individual attempts to resolve fail, the robots enter into a cooperative phase of resolving the conflict at the second level. At the third level, conflicts are propagated to robots not directly involved in any conflicts at that instant and their assistance sought for possible resolution. At all the three levels conflict resolution is attempted by velocity control of the robots through an algorithm that is distributed across all the robots in the system. Simulation results are shown to validate the efficacy of this methodology.