PhD - Predictive Occupancy World Models for End-to-End Autonomous Driving

Develop a generic predictive occupancy model as the core representation of the driving scene, designing architectures that capture the geometry and semantics of all objects to enable an explicit understanding of 4D world dynamics. Research, implement, and evaluate novel end-to-end driving architectures based on these models to ensure robust situational awareness and safe navigation. Create state-of-the-art perception systems that transform raw sensor data into rich spatial representations and driving actions, where decisions are directly based on predictive occupancy world models. Rigorously test your models in simulation and with real-world data, demonstrating improvements in safety, transparency, and performance, and publish your findings at leading AI and robotics conferences. The final PhD topic is subject to your university. Start: September 2026.