Home > About > MSc Thesis Projects

Looking for a Challenging MSc Thesis Project on AI in Mobility?

No need to look any further! AiMTT and TU Delft have created several projects that offer an excellent opportunity for students to work on real-life traffic challenges using cutting-edge AI techniques.

Our Msc Thesis Projects:

What Makes a Good Driver? Learning and Inference in Traffic

How do drivers decide when to accelerate, brake, or keep their distance? Traditional traffic models describe this using fixed rules, but Active Inference offers a new perspective: drivers as intelligent agents who form beliefs about their surroundings and act based on predictions of what will happen next. This project explores a central question: Is it really possible to model drivers as adaptive, belief-driven agents, and what does this mean for traffic flow?

Download more information…

Driving with Limited Information: What Can Drivers Really See?

Modern mobility systems generate a constant stream of data. Turning these data into reliable short-term forecasts is essential for operations, customer information, and network management. While classical time-series models remain strong baselines, today’s complex and dynamic networks increasingly benefit from so-called Long Short-Term Memory models. These models can remember what parts of recent history still matter.

Download more information…

From Reaction to Anticipation: How Can Drivers Predict the Future?

To make sense of the complex, living webs of urban mobility, we need technology that can “see” the intricate geometry of a city. Graph Neural Networks (GNNs) are models designed to master exactly these complexities. But what exactly is a “graph” in the world of AI? How do these models pass information between different parts of the network? And how can they support transportation?

Download more information…

Can Traffic Self-Organise? Decentralised Control with Intelligent Agents

Most traffic systems are managed through centralised control, such as coordinated control strategies. Such approaches can be difficult to scale and adapt, especially in highly dynamic or uncertain environments. At the same time, traffic is fundamentally a multi-agent system in which many independent actors interact locally. This raises an alternative perspective: instead of controlling traffic from the top down, can we design rules that allow traffic to organise itself from the bottom up?

Download more information…


Our MSc thesis projects are, in principle, open to TU Delft students only. If you are interested in one of these projects but are not a TU Delft student, please send us an email.