Predictive AI-based Traffic Control - Germany

The project ‘Predictive AI-based Traffic Control’ of the University of Applied Sciences Konstanz in cooperation with the City of Konstanz, explores inter alia motion intentions of traffic participants (pedestrians, persons with mobility aids, crossing cyclists) at signal-controlled intersections. The system uses roadside Light Detection And Ranging (LiDAR) sensors, which are pole-mounted at road intersections.  Based on the motion intentions (‘predictions’), the system activates the pedestrian phase of the traffic lights.

The LiDAR system produces virtual point-clouds; AI detects and tracks motion patterns out of them. Since June 2025, AI collects anonymised data under differing weather and lighting conditions in two pedestrian-activated traffic lights of Konstanz. AI learns the shape of pedestrians in the point-cloud, the use of mobility aids and the realisation of a presumed crossing intention. The LiDAR system will be operated and tested in the real life during the 2026 summer period.

The detection of the traffic participants occurs locally without any cloud connectivity. The system operates with a power as low as 100 watt and detects the pedestrians in real-time out of the point-cloud, predicts accurately the motion patterns and distinguishes reliably the subgroups of pedestrians.  The system recognises the prospective crossing of traffic participants in a time horizon of (up to) 10 seconds. The pedestrian phase is then activated without delay and just when necessary. The controller dynamically adapts crossing time and may activate audible signals. The touchless system makes the manual use of push-buttons obsolete and reduces the waiting time until the actuation of the pedestrian phase.

Learn more about the project ‘Predictive AI-based Traffic Control’.

Disclaimer:
Accessible EU is not responsible for the accessibility of content provided by third parties.