Physics-informed 3D World Model in Autonomous Driving

Institut

Professur für autonome Fahrzeugsysteme

Typ

Semesterarbeit / Masterarbeit / HiWi-Tätigkeit /

Inhalt

 

Beschreibung

Modern autonomous vehicles are increasingly reliant on advanced perception systems to navigate complex environments. However, these systems often lack a deep understanding of the underlying physical dynamics of the world, which can limit their ability to make safe and reliable driving decisions. To address this, we propose the development of a physics-informed 3D world model for autonomous driving.

At the AVS Lab, we are pioneering a novel approach that integrates physical laws and constraints into the 3D modeling of the driving environment. This model will enhance the vehicle's ability to predict and respond to dynamic changes in the environment, such as moving obstacles and varying road conditions.

Project Objective:

This project aims to develop a comprehensive 3D world model that incorporates physical principles to improve the decision-making process in autonomous driving. Key components include:

• Developing a physics-informed framework to model dynamic interactions within the driving environment.
• Integrating this framework with existing perception systems to enhance scene understanding and prediction accuracy.
• Evaluating the model's robustness in diverse scenarios.

We offer:

• Cutting-edge research at the intersection of physics, perception, and autonomous driving.
• Collaboration with a multidisciplinary team and integration with an existing autonomous driving stack.
• Opportunities to publish in top-tier robotics and AI conferences.
• A supportive, English-speaking research environment with potential for thesis supervision.

Your qualifications:

• Strong programming skills in Python and experience with machine learning frameworks such as PyTorch.
• Interest in autonomous driving, physics-based modeling, and deep learning.
• A solid background in computer vision and robotics is advantageous.
• You are not afraid of mathematical proofs concerning diffusion.

The start date is flexible. If you're interested, please send:

• A brief description of your relevant coursework and experience in computer science, particularly in deep learning and coding (200 words).
• Academic transcript (optional for now, but may be required later).
• CV (optional).

Please send your application to dingrui.wangtum.de.

Möglicher Beginn

sofort

Kontakt

Dingrui Wang
dingrui.wangtum.de