Game theoretic model for lane changing: Incorporating conflict risks

Abstract

The study employs a Quantal Response Equilibrium framework to model lane changing manoeuvres. Prior game theoretic studies in lane changing have pre-eminently assumed Nash equilibrium solutions with deterministic payoffs for actions. The study method involves developing expected utility models for drivers’ merge and give-way decisions. These utility models incorporate explanatory variables representing driver trajectories during a lane changing manoeuvre. The model parameters are estimated using maximum likelihood on lane changing data at a freeway on-ramp using the NGSIM dataset. Based on the estimated parameters it was concluded that longer acceleration lanes and reduction of speed limits on on-ramps could help significantly reduce likelihood of conflict. To demonstrate the robustness of the model, predictions of lane changing on an out-of-sample data were found to be reasonably accurate.