Reliable identification and prediction of traffic situations are crucial for intelligent traffic control and active traffic management. With static traffic lights, for example, green times are designed around a best possible compromise for all relevant traffic situations. When adaptive traffic control schemes are in place, optimal green times for each relevant traffic situation have to be found. What are the 'relevant' traffic situations and how can they be identified? Too often, misidentification or consideration of non-relevant situations at the design stage cause sub-optimal green times or unnecessary system complexity. However, researchers and engineers at Andata - an Austrian R&D company - has developed and implemented a new, self-learning method, which can automatically and autonomously identify the relevant traffic situations of particular streets, crossings, road networks and even complete city sections. This is done by considering all kinds of traffic information, from floating car data (FCD), to static traffic sensors, to virtual sensors, or a combination of any of these. The information is put together using special metrics, before artificial intelligence is used to automatically cluster and sort it into a catalog of real-life scenarios that occur at the investigated region.
展开▼
