Treecoding for Scalable Belief Exchange in Autonomous Vehicle Control Systems

Abstract

Treecoding techniques originally developed for ecient simulation of large physical systems are applied to the domain of autonomous vehicle control systems to significantly improve the scalability of inter-vehicle communication requirements. The autonomous control system in question requires that cooperating vehicles periodically broadcast their knowledge about relevant features of the environment (“beliefs”). For a group of M vehicles each with N beliefs, this results in required communications bandwidth scaling as O(MN). Use of treecoding improves the scaling to O(M logN), allowing the autonomous control system to function in much more complex environments. The principles of the autonomous control system and the treecoding technique are reviewed, a treecode-based information exchange algorithm is described, and simulation results are presented for an area-search scenario showing the eect on communication bandwidth and mission performance.