Monday, July 21, 2025

AI for Automated Driving: End-to-End Driving Models meet Connectivity

Topic and Goal of the Thesis

Artificial Intelligence plays an increasing role in automated driving systems. Big players like Tesla, Wayve, Waymo and others are already exploring the option to replace almost the complete driving stack with one neural network, a so-called end-to-end driving model.

This thesis aims to bring these models to the next level. It explores their potential to take as input additional data shared by other vehicles and/or smart roadside infrastructure (V2X data). Connectivity allows these E2E models to make better decisions, hence making traffic safer.

Working Points

• Research of relevant literature
• Assessment of available E2E models and training pipelines
• Extension of E2E models with the ability to process V2X data
• Training of E2E models
• Open-loop and closed-loop evaluation
• Optional: Demonstration in our real-world vehicle

Requirements

• Reliability, commitment and enjoyment of working independently
• Experience with Python or C++
• Experience with the following is beneficial: ROS, Docker und Machine Learning

Note: Please attach brief resume and grade summary.