Automotive Collision Avoidance Optimization with Probabilistic Traffic Models

As the automotive industry moves towards autonomous driving, it becomes increasingly necessary to develop advanced collision avoidance systems, crash prediction systems, and the tools for their rigorous analysis. Probabilistic traffic models, providing a statistical representation of the future behavior of traffic participants, are crucial for risk estimation in automotive collision avoidance systems, and can be used in both optimal policy generation and in safety system validation. This work investigates optimal policy generation for autonomous driving safety systems in which the driving problem is formulated as a Markov decision process. Optimal solutions are found by leveraging traditional solution techniques.