SPP 1835: Kooperativ interagierende Automobile/en
Situation assessment and semantic maneuver planning under consideration of uncertainties for cooperative vehicles
An automated, cooperative car must take decisions in a highly dynamic, interacting and only partially observable environment. So far, existing approaches are limited by considering situations only from an egocentric perspective and without considering cooperation aspects with and between other road users. Moreover, due to the high complexity in general only few, often predefined situation aspects are considered.In this project, probabilistic methods for situation awareness and -prediction, as well as the cooperative behavior decision within real traffic and with explicit consideration of cooperation are explored. These are based on the results of the cooperative perception and give input to the cooperative trajectory planning.The inference of relations and interactions between road users with different cooperation levels, in heterogeneous traffic, is addressed by an object-oriented probabilistic relational approach. Especially by considering implicit cooperation the situation assessment will lead to a reasonable behavior. To anticipate the consequences of own maneuvers under uncertainty and cooperation, the cooperative behavior decision has to be formulated as an extended markov decision process. The definition of the planning space as an abstract, semantic space, reduces the complexity for the planning algorithm.For evaluation, a new simulation method will be explored. It automatically derives behaviors of cooperative road users from existing sensor records and synthesis them according to the behavior of the test vehicle.