Masterarbeit Sensor- Gaussian Scene Representations Simulation in Autonomous Driving

This project investigates how neural scene representations based on 3D Gaussian Splatting can be extended to support simulation of sensor data in autonomous driving. The main goal is to design a sensor‑ready Gaussian representation that encodes not only geometry and appearance, but also semantic and sensor‑relevant properties required for simulating different sensing modalities. Starting from recorded sensor data, scenes shall be reconstructed and enriched with these attributes to enable simulation of different sensor types in the future.

Rather than focusing on highly accurate sensor models, the project emphasizes the representation itself as a foundation for multimodal simulation.

Key Questions

1. What are the limitations of existing Gaussian‑based scene representations for sensor simulation, and how can they be extended to become sensor‑ready?

2. Which semantic and physical attributes must be encoded in Gaussian primitives to enable flexible simulation of different sensor modalities?

3. How well can sensor‑ready Gaussian representations support re‑simulation of sensor data compared to real‑world observations?

Qualifications

• Computer Science

• Robotics

• Math, Data Science or comparable degree program

• Machine learning / deep learning

• Digital image processing

• Solid understanding of popular machine learning and deep learning concepts

• Experience with computer vision

• Experience with Gaussian splatting is a bonus

• Confident use of Microsoft Office, Git, and Linux (Ubuntu)

• In‑depth knowledge of Python, C, or C++

• Proven experience with popular machine learning frameworks

• English (fluent spoken and written)

• German is an advantage

Soft Skills

• High level of initiative

• Strong analytical skills

• Structured and independent way of working

#J-18808-Ljbffr