A Multiclass Decision-Making Framework for Controlling Autonomous and Conventional Vehicles at Intersections

Current research mostly focuses on developing rules and protocols to improve the safety and efficiency of intersections through autonomous vehicles. These scholars always deal with the era of fully automated vehicles and smart infrastructures. However, there will be a sufficient transition period during which smart and conventional vehicles share the same transportation networks and intersections. In this case, the ability to improve intersection efficiency will be ineffective. The autonomous vehicle cannot predict the next human behavior; hence the current intersection control rules may not change. This paper proposes a simple decision-making methodology to control the potential conflict between two autonomous and conventional vehicles at intersections. The methodology applies the Perfect Bayesian Equilibrium game theory to predict human behavior and enhance the autonomous vehicle’s decision strategy. The results showed that the autonomous vehicle’s decision is influenced by the driver’s awareness level of the possible conflict and the available sight distance at the intersection. The results also showed an increase in the safety and efficiency of intersections compared to the current conditions.