AI-based optimization of pedestrian crossing times by means of smart traffic signals – KI4PED

Project description

Current studies indicate that the green phases for pedestrians at traffic signals do not take the time into consideration that different categories of people require to cross the road. It also appears that the duration of the green phase influences the readiness to cross the road during a red phase. Intelligent approaches aiming at an early detection of pedestrians allow for optimized traffic signal control, thus reducing waiting times and increasing road safety.

Outcome and effects

The KI4PED project developed an innovative approach to a demand-responsive control of traffic signals for pedestrians which consists of a combination of 3D-LiDAR sensors and AI-based data evaluation. This allows for a contactless demand-based activation for pedestrians and reduces waiting times for them. Another objective of this approach is to reduce jaywalking during the red phase. The chosen approach is largely robust against lighting influences and respectful of privacy rights thanks to a GDPR-compliant overall system approach. At selected traffic signal locations, data was recorded with 3D-LiDAR sensors and used as a basis for developing an AI-based prediction of whether individuals intend to cross the road. Furthermore, the possibility to integrate the system into existing traffic light controls was examined and a corresponding control integration concept was developed. As part of the evaluation, the system was examined for its suitability for smart traffic light control at the selected locations and for its robustness against various lighting conditions.

Project coordinator	Fraunhofer IOSB-INA, Lemgo • Funding code: 19F1090A
Project volume (at the time of approval)	81,364.00 euro (95 % of which are funded by the BMDV)
Project duration (at the time of approval)	10/2021 – 09/2022
Contact	Fraunhofer IOSB-INA Dr.-Ing. Dennis Sprute +49 (0)5261-9429063 dennis.sprute@iosb-ina.fraunhofer.de