Advances in Dynamic Games

Volume 11 of the series Annals of the International Society of Dynamic Games pp 253-271


Collision Avoidance Strategies for a Three-Player Game

  • Sriram ShankaranAffiliated withGeneral Electric Research Center Email author 
  • , Dušan M. StipanovićAffiliated withDepartment of Industrial and Enterprise Systems Engineering, and Coordinated Science Laboratory, University of Illinois at Urbana-Champaign
  • , Claire J. TomlinAffiliated withDepartment of Electrical Engineering and Computer Sciences, University of California

* Final gross prices may vary according to local VAT.

Get Access


Collision avoidance strategies for a game with three players, two pursuers and one evader, are constructed by determining the semipermeable curves that form the barrier. The vehicles are assumed to have the same capabilities, speed, and turn-rates. The game is assumed to be played on a two-dimensional plane. We consider avoidance strategies for a particular form of the game defined in the following way: the pursuers are assumed to act noncooperatively, the evader upon realizing that one (or both) of the pursuers can cause capture, takes an evasive action. We find states from which the pursuer can cause capture following this evasive action by the evader. The envelope of states that can lead to capture is denoted by the barrier set. Capture is assumed to have occurred when one (or both) pursuers have reached within a circle of radius, l, from the evader. The usable part and its boundary are first determined along with the strategy along the boundary. Semipermeable curves are evolved from the boundary. If two curves intersect (they have a common point), the curves are not extended beyond the intersection point. As in the game of two cars, universal curves and the characteristics that terminate and emanate from the universal curve are used to fill voids on the barrier surface. For the particular game (and associated strategies) considered in this paper, numerical simulations suggest that the enlarged set of initial states that lead to capture is closed. As the game considered here is a subset of the more complete game, when two pursuers try to cause capture of a single evader, the avoidance strategies are most likely to belong to the set of strategies for the complete game.