Predictive Traffic Light Control with LiDAR-Based Intent Prediction

Sun, 12 May 2024 11:13:50 +0100

The project Predictive AI-Based Traffic Light Control aims to improve the interaction between pedestrians, cyclists and drivers at intersections by anticipating the future paths of road users and their expected duration. LiDAR sensors are planned for this purpose, as they allow the intentions of pedestrians and cyclists to be detected and predicted in a privacy-compliant way. The goal is to control conventional demand-actuated traffic lights automatically and without manual push-button activation, taking into account traffic flow, pedestrian waiting times and vehicle idle times.

The system can also be used to prioritize cyclists under certain conditions, for example when it is raining. In addition, the algorithms developed in the project are intended to warn cyclists of hazards at critical and confusing locations. This covers hazards caused by other road users as well as by inappropriate riding behavior of the cyclists themselves.

The approach combines methods from artificial intelligence with exteroceptive sensors. Analyzing the sensor data and predicting the behavior of road users allows the traffic light phases to be adapted to the needs of all road users.

More information about the Smart Green City initiative is available (in German).

The project page of the predictive traffic light control can be found (in German).

LiDAR | Christopher Knievel

Predictive Traffic Light Control with LiDAR-Based Intent Prediction