Modular Decision Making Framework for Level 4 Applications in Automated Driving

Decision making for automated driving according to SAE level 4 (L4) faces a major challenge in the urban environment. Previous research has presented a variety of approaches, which, however, mostly focus on decision making for individual maneuvers or use cases. To address the complexity and requirements of the urban environment, it seems useful to present a way to flexibly combine approaches from the literature. Addressing this challenge, we present a framework for decision making that allows a modular embedding of different methods and thus scalability. The framework includes all necessary processing steps and submodules within a functional architecture according to SAE level 4, starting from the environment model up to the trajectory planning. The central component is formed by automatically generated interaction areas, which serve as the basis for the modular structure of the downstream processing levels. The universal applicability of the framework is demonstrated by real implementations in three representative L4 functionalities.