Driver Modeling through Deep Reinforcement Learning and Behavioral Game Theory

In this paper, a synergistic combination of deep reinforcement learning and\nhierarchical game theory is proposed as a modeling framework for behavioral\npredictions of drivers in highway driving scenarios. The need for a modeling\nframework that can address multiple human-human and human-automation\ninteractions, where all the agents can be modeled as decision makers\nsimultaneously, is the main motivation behind this work. Such a modeling\nframework may be utilized for the validation and verification of autonomous\nvehicles: It is estimated that for an autonomous vehicle to reach the same\nsafety level of cars with drivers, millions of miles of driving tests are\nrequired. The modeling framework presented in this paper may be used in a\nhigh-fidelity traffic simulator consisting of multiple human decision makers to\nreduce the time and effort spent for testing by allowing safe and quick\nassessment of self-driving algorithms. To demonstrate the fidelity of the\nproposed modeling framework, game theoretical driver models are compared with\nreal human driver behavior patterns extracted from traffic data.\n