Title: Game theory based autonomous vehicles operation
Authors: Changwon Kim; Reza Langari
Addresses: Department of Mechanical Engineering, Texas A&M University, College Station, TX 77843-3123, USA ' Department of Mechanical Engineering, Texas A&M University, College Station, TX 77843-3123, USA
Abstract: In this paper, we propose a game theory based approach to decision making with application to the operation of autonomous ground vehicles in highway setting. The mixed-motive game theory is utilised as a decision-making strategy in the context of a two-player game involving autonomous vehicles. The payoff matrices are defined by considering the safety of each players decision combination in view of their desire to stay within a given lane or to change lanes in consideration of the traffic conditions that the vehicles encounter. By analysing the payoff matrix, either a pure (deterministic) strategy or a mixed (probabilistic) strategy is selected. Three 10 km velocity profiles are predefined for simulation purposes. The simulation results demonstrate effective driving performance. In particular when it is compared with non-game theory cases, game theory based results show larger payoff for both vehicles and smaller payoff differences, securing safe manoeuvring via lane change manoeuvre (LCM) and adaptive cruise control (ACC).
Keywords: autonomous vehicles; game theory; mixed motive games; virtual bumper; decision making; vehicle safety; lane changing; velocity profiles; simulation; driving performance; safe manoeuvring; lane change manoeuvres; adaptive cruise control; ACC.
International Journal of Vehicle Design, 2014 Vol.65 No.4, pp.360 - 383
Received: 26 Apr 2012
Accepted: 12 Mar 2013
Published online: 30 Oct 2014 *