AI-controlled traffic lights tested in Lemgo

Researchers at the Fraunhofer Institute for Optronics, System Technologies and Image Exploitation IOSB, specifically its Industrial Automation branch (IOSB-INA) in Lemgo, are developing an artificial intelligence system for traffic light control. The projects, named KI4LSA and KI4PED, aim to improve traffic flow, reduce waiting times, and increase safety for pedestrians at intersections.

Traditional traffic light systems rely on rigid, rule-based controls that often fail to adapt to real-world traffic conditions. Current sensors, such as inductive loops embedded in the road surface, provide only a basic overview of traffic. The Fraunhofer team is utilizing more advanced sensors, including high-resolution cameras and radar, to gather precise, real-time data on vehicle counts, speeds, and queue lengths.

This data is fed into AI algorithms employing deep reinforcement learning. The system analyzes traffic situations and calculates optimal signal timings to manage vehicle and pedestrian movement. The objective is to minimize congestion, thereby reducing associated noise and carbon dioxide emissions. The KI4PED project specifically focuses on using LiDAR sensors and AI to detect and track pedestrians for improved crossing safety.

Testing has been conducted in a simulated traffic environment based on the city of Lemgo. The AI has been trained on extensive datasets derived from actual rush-hour traffic, allowing it to learn optimal control strategies for complex scenarios. The system learns to identify the best ways to manage traffic flow through numerous iterations within the simulation model.

The KI4LSA project is funded by the German Federal Ministry of Transport and Digital Infrastructure (BMVI) and is scheduled to conclude in summer 2022. Project partners include Stührenberg GmbH, Cichon Automatisierungstechnik GmbH, Stadtwerke Lemgo, and Straßen.NRW.