A Game Theory Based Model of Human Driving with Application to Autonomous and Mixed Driving
Abstract
In this work, I consider the development of a driver model to better understand human drivers’ various behaviors in the upcoming mixed situation of human drivers and autonomous vehicles. For this, my current effort focuses on modeling the driver’s decisions and corresponding driving behaviors.
First, I study an individual driver’s reasoning process through game theoretic investigation. The driver decision model is modeled as the Stackelberg game, which is based on the backward information propagation. In the driver decision model, I focus on the drivers’ insensible desires and corresponding unwanted traffic situations. With the comparison of the model and the field data, it is shown that the model reproduces the relationship between the driver’s inattentiveness and collisions in the real world.
Next, the driving behavior control is presented. I propose a human-like predictive perception model of potential collision with an adjacent vehicle. The model is based on hybrid systematic approach. In turn, with the predictive perceptions, a driving safety controller is designed based on model predictive control. The model shows adequate predictive responses against the other vehicles with respect to the driver’s rationality.
In sum, I present a driver model that corresponds to and predicts traffic situations according to a human driver’s irrationality factor. This model shows a meaningful similarity to the real-world crashes and predictive behaviors according to the driver’s irrationality.
Citation
Yoo, Je Hong (2014). A Game Theory Based Model of Human Driving with Application to Autonomous and Mixed Driving. Doctoral dissertation, Texas A & M University. Available electronically from https : / /hdl .handle .net /1969 .1 /153641.