Ethics-Aware Decision-Making with Large Language Models

Institut

Professur für autonome Fahrzeugsysteme

Typ

Bachelorarbeit / Semesterarbeit /

Inhalt

experimentell / theoretisch /  

Beschreibung

Background

The autonomy of vehicles has advanced rapidly in recent years, reaching a level where human intervention is barely or not at all required in certain controlled environments. Leading the way are manufacturers that now offer Level 3 autonomous vehicles depending on the system’s design. This progress relies heavily on the development and validation of highly reliable driving functions. Ensuring their safety and reliability requires extensive testing in diverse and challenging scenarios.

In parallel, the rise of Large Language Models (LLMs) has opened new opportunities for autonomous driving research. LLMs combine strong generalization with adaptability, enabling direct application to domain-specific tasks such as perception, decision-making, control, and simulation. Recent studies highlight their integration into modular and end-to-end driving systems. However, existing work has primarily focused on the broader scope of autonomous driving and has paid limited attention to explicit decision-making under ethical principles. This leaves open an important research direction: exploring how LLMs can be leveraged not only for task automation but also for embedding ethics-aware decision logic into simple but safety-relevant choices such as whether to change lanes, yield, or accelerate.

---

Objective

The primary objective of this project is to develop a framework for ethics-conditioned decision-making using Large Language Models (LLMs) and to evaluate their principle adherence and counter-reasoning robustness.

The framework will consist of two key components:

1. Ethics-conditioned decision selector:
   A set of candidate decisions (e.g., change lane, keep lane, brake, accelerate) is generated along with per-agent risk metrics.

2. Validation & robustness module:
   An automated verifier checks whether the LLM’s choice is optimal for the declared principle and runs counter-reasoning tests (metric–text conflicts, irrelevant framing, perturbation invariance, and self-critique re-decide). The resulting decisions are used to evaluate and improve the robustness of ethics-aware reasoning in autonomous driving.

---

We Offer

• A dynamic and future-oriented research environment

• Hands-on experience with a state-of-the-art software stack for autonomous driving (CARLA/CommonRoad)

• Opportunity to publish a scientific paper (based on merit)

• The thesis can be written in either English or German

---

Requirements (What You Should Bring)

• Initiative and a creative, problem-solving mindset

• Excellent English or German proficiency

• Advanced knowledge of Python

• Prior experience with autonomous vehicles or Large Language Models is an advantage

• Familiarity with common software development tools (e.g., Git, Ubuntu) is desirable

Work can begin immediately.
If you are interested in this topic, please first have a look at our recent survey paper: https://arxiv.org/abs/2506.11526
Then send an email with a brief cover letter explaining why you are fascinated by this subject, along with a current transcript of records and your resume, to: yuan_avs.gao@tum.de

Tags

AVS Gao

Möglicher Beginn

sofort

Kontakt

Yuan Gao
yuan_avs.gaotum.de