A Game-Theoretic Approach to Co-operative Context-Aware Driving with Partially Random Behavior

We present a game-theoretic approach to co-operative driving tinder the assumption that drivers are willing, yet unable to follow recommendations of their adaptive driving assistance systems (ADAS) precisely. Unknown road conditions, certain health conditions, or other external impacts may render drivers unable to exactly realize their optimal behavior. We argue the importance of context information for co-operative driving, and sketch a framework where driving assistance systems communicate with each other to negotiate optimal recommendations for their specific drivers. As the final decision is nevertheless up to the driver, it is interesting to investigate what happens if drivers are unwilling or unable to follow the recommended choices. Using game theoretic tools, we derive bounds on the impact that partially random behavior can have on the own utility (safety) of a driver.