Human cognition comprises perception, reasoning and insight; their technical correspondence uncloses entirly new skills. A cognitive automobile perceives itself and his environment self-reliant with different sensors as well as through communication with other traffic participants. It structures the perceived information and is capable to safely and efficiently navigate in urban traffic due to his knowledge.
The goal of the research group "cognitive automobiles" is to give automobiles an abstract understanding of traffic situations and to make autonomous behavior decisions on this basis. We intend to create a consistent probabilisitic model of the vehicles knowledge about its environment and itself. This is necessesary since a vehicle can only oberserve its sourroundings partially and noisy. It is therefor of crucial importance that the vehicle is aware of this uncertainties and that this knowledge flows into the behavior decision process.
An additional focus lies in the application of machine learning methods to make the vehicle more intelligent. The automatic reasoning about the behaviors of other traffic participants as well as the risk potential of situations and relevant aspects for the decision provides not only generalizing methods but also more realistic heuristics than manual design. An important data source constitutes the human background knowledge incorporated by a driving instructor.