Can AI Learn How Traffic Lights Think? | Diego Van Overberghe (ETH Zürich) | Colloquium #08

🎓 TU Delft | Delft Center for Systems and Control (DCSC) 🇳🇱
📚 Colloquia Series – Recording #08

Traffic signals play a crucial role in shaping urban mobility, yet accurately reproducing their behaviour in microscopic traffic simulations remains a major challenge. In this DCSC colloquium, we explore how machine learning techniques can be used to uncover and predict the logic behind real-world traffic signal controllers. 🚦🚗

Welcome to the official recordings of the DCSC Colloquia Series at TU Delft!
In this series, we share insightful talks from leading researchers presenting their latest work in systems, control, optimization, and learning.
🔗 https://www.youtube.com/redirect?event=video_description&redir_token=QUFFLUhqbGZKbWZLS1VkMGNnX25mRFdqSEhaWG52bVZSd3xBQ3Jtc0trMjZKejlpUzROMlhwVmFHNGxxZjFqMm9SVXNFUy1VNzJqcXlVTXhqdGdCUW5YRExldS1XU3JaT0dQLTI0LUtLX0E1dkhleTRCdGpBb2U2TWRZbm5uTzBuaE1yRlJaMFQ0NEEzTkpuekE3ME1vUGRqcw&q=https%3A%2F%2Fwww.tudelft.nl%2Fme%2Fover%2Fafdelingen%2Fdelft-center-for-systems-and-control&v=8e3IfivcMQw

⸻

🎥 Talk Title:
Learning to Predict Traffic Signals

📣 Speaker:
Diego Van Overberghe (ETH Zürich, Institut für Automatik – IfA)
LinkedIn: https://www.linkedin.com/in/diegovano/

⸻

🧠 Abstract:
Microscopic traffic simulations are essential tools for evaluating and improving urban traffic infrastructure. Since traffic flow in cities is strongly influenced by traffic signals and their programming, simulations must accurately reproduce signal behaviour and operation.

This talk presents a decision-tree-based approach for learning the underlying logic of traffic signal controllers directly from data. Unlike methods relying on fixed schedules or regression-based prediction of future signal timings, the proposed framework aims to recover the controller logic itself.

The method is first validated on a synthetic scenario, where it successfully reconstructs the traffic signal controller behaviour. The approach is then applied to a real intersection in Zürich, demonstrating the ability to capture key operational behaviours such as public transport prioritisation — despite having no prior knowledge of the controller implementation.

⸻

In this talk:

🚦 Learning traffic signal controller logic
🧠 Decision-tree-based modelling approaches
🏙️ Microscopic urban traffic simulation
🚌 Public transport prioritisation in signal control
📊 Data-driven modelling of traffic infrastructure

⸻

🔬 Research Cluster:
Modelling and System Identification

⸻

👤 About the Speaker:
Diego Van Overberghe is currently a Research Assistant at the Automatic Control Institute (IfA) at ETH Zürich.

His research focuses on developing platforms and tools to facilitate traffic research using SUMO (Simulation of Urban MObility), a widely used microscopic traffic simulator. His work builds on experience gained during his Master’s thesis project in traffic simulation and control.

⸻

📈 Keywords:
traffic signals, traffic signal control, microscopic traffic simulation, SUMO, intelligent transportation systems, urban mobility, traffic modelling, system identification, machine learning for traffic systems, decision trees, public transport prioritisation, ETH Zürich, systems and control, DCSC

Hashtags:
#TrafficEngineering #UrbanMobility #TrafficSimulation #IntelligentTransportationSystems #MachineLearning #SystemIdentification #SUMO #ControlTheory #TUDelft #DCSC

Видео Can AI Learn How Traffic Lights Think? | Diego Van Overberghe (ETH Zürich) | Colloquium #08 канала DCSC TU Delft