Event Discrete Decision-Making for Autonomous Vehicles at Non-Signalized Intersections

Non-signalized urban intersections are one of the remaining challenging situations when it comes to autonomous driving. Situations may arise that are not clearly regulated by the traffic regulations and therefore require cooperation between vehicles to resolve the situations. This paper introduces a general modeling approach for driving at intersections and decision-making there. The decision process is modeled based on discrete event systems (DES). Within the modeling, limited visibility is considered. Unexpected behavior of other vehicles as well as deadlock situations are also accounted for. The decision-making process is based on explicitly modeled features and events and can thus be explained to and understood by its users.

The algorithm is evaluated with extensive simulations and the results show that the model is able to deal with multiple cooperation vehicles and reliably avoids collisions even during unexpected behavior of the cooperation vehicles and deadlock situations.