Edinburgh Urban Dataset

Designing shared autonomy systems that intelligently integrate human intent with safe AI control.

The Assistive Driving Mission focuses on making driving safer and more supportive by rethinking the role of AI in the car. Rather than replacing the driver, our approach is to build systems that work alongside people—understanding their attention, intentions, and capabilities in real time, and providing the right level of assistance when it matters most. This is especially important in today’s roads, where traffic conditions are increasingly complex and drivers vary widely in experience, confidence, and cognitive load. By developing AI that can collaborate with drivers—offering guidance, sharing control, or intervening in high-risk situations—we aim to create a new generation of driving technology that is more adaptive, trustworthy, and human-centred.

A key part of this work is the creation of a large-scale urban driving dataset, captured here in Edinburgh. Unlike traditional driving datasets that focus mainly on detecting objects (like cars or pedestrians), our dataset is designed to capture the rich interactions that happen on real streets—how drivers, cyclists, and pedestrians behave in relation to one another, and how risk emerges in complex scenarios. This includes detailed, automatically annotated data that helps AI systems learn not just what is happening, but why it matters for safety and decision-making. Building this dataset is essential because current AI systems struggle with these interaction-heavy, real-world situations. By providing high-quality, realistic data, we enable new kinds of intelligent systems that can better understand context, anticipate hazards, and support drivers in a more meaningful way—ultimately helping to reduce accidents and improve everyday driving experiences.