Borchers, PhilippPhilippBorchersHagemann, WillemWillemHagemannGrundt, DominikDominikGrundtWerner, TinoTinoWernerMüller, JulianJulianMüller2025-01-282025-01-282024https://publica.fraunhofer.de/handle/publica/48299910.1007/978-3-031-66428-1_12In recent years, the integration of knowledge into AI training processes has been shown as a promising approach to improve AI performance, training costs and resource efficiency. Here, the formalization of knowledge is a key challenge. In this article, we discuss the capabilities of the visual yet formal specification language called Traffic Sequence Charts (TSC) on formalizing multimodal knowledge, in particular procedural knowledge about traffic maneuvers. Finally, we present an approach using the formalized knowledge to train reinforcement learning (RL) agents, aiming to transform the descriptive knowledge on traffic maneuvers in TSCs into performative knowledge in AI traffic agents. To this end, we were able to train an agent to control a vehicle through a pass-by maneuver and apply it successfully to an untrained overtaking maneuver.enAbstract scenario specificationKnowledge formalizationReinforcement learningTraffic scenariosconference paper