Development of a Predictive Collision Avoidance for Subjective Adjacent Risk Estimation

In this paper, we have considered the development of predictive collision avoidance as an individual driver behavior modeling. First, our investigation focuses on developing a subjective collision risk estimation model. Through the game theoretic estimation of the counterpart's behaviors and the corresponding time-evolution of the unsafe collision areas, we compute an objective collision model. In turn, we design a human-like predictive perception model for the collision with an adjacent vehicle, based on the objective collision model and the driver's subjective level of safety assurance. Next, a driving controller is designed to optimally avoid the anticipated collision for the prediction time horizon using model predictive control, which is founded on the subjective collision estimate that varies for every individual who has different aggressiveness. Simulation results indicate that the subject vehicle can react to the surrounding vehicles even without immediate actions from the counterpart. This simulates a typical driver's reasoning in view of his/her disposition so that the driver's reaction in response to roadway traffic is appropriately considered.