Asynchronous real-time framework for knowledge-based intersection assistance

Recently, a demand for advanced driver assistance in sophisticated situations with multiple traffic objects and complex infrastructure has emerged. To handle this kind of situations we developed a knowledge-base to model abstract qualitative information of traffic situations, up to very complex traffic intersections. We further created an asynchronous real-time simulation framework to investigate applicability of knowledge-based driver assistance systems within vehicle or traffic management systems. While the simulation with sensor data is running with very short update cycles the knowledge-base is updated asynchronously since reasoning is time expensive. Driver assistance functions are able to query the knowledge-base once it is fully reasoned. We especially focus on safety assistance systems and performed tests of our knowledge-based framework on exemplary intersection assistance functions. Results show the capability to perform semi and fully autonomous warning and deescalation assistance functions in real-time.