[MA/BA/SA] Robust trajectory planning via Multi-Modal Sensor Fusion

22.05.2026, Abschlussarbeiten, Bachelor- und Masterarbeiten

This thesis aims to develop a multi-modal sensor fusion framework for robust trajectory planning in autonomous driving using multi-modal sensors in the simulation environment.

Autonomous driving systems increasingly rely on multi-modal perception to achieve robust and safe navigation in complex real-world environments. While recent end-to-end driving approaches have demonstrated promising results using mixture of sensors, many existing systems still struggle to generalize under challenging conditions such as adverse weather, low-friction roads, sensor occlusions, and dynamic traffic scenarios. Furthermore, purely vision-based approaches often lack the geometric understanding and robustness required for reliable trajectory planning.
This thesis aims to develop a multi-modal sensor fusion framework for robust trajectory planning in autonomous driving using multi-modal sensors in the simulation environment. The goal is to investigate modern fusion architectures for end-to-end neural planning and evaluate their robustness, generalization capability, and planning performance under diverse driving conditions.
The project is part of an EU Software-Defined Vehicle (SDV) research initiative focusing on next-generation AI-driven vehicle architectures and intelligent driving systems.

Objectives:
- Investigate state-of-the-art multi-modal fusion architectures for autonomous driving, including transformer-based and end-to-end neural planning approaches.
- Design and implement a sensor fusion pipeline in the dedicated simulation environment.
- Explore different fusion strategies.
- Evaluate the robustness and generalization capability of the learned representations under varying environmental and traffic conditions.
- (Optional) Investigate the integration of Large Language Models (LLMs) or Vision-Language Models (VLMs) for scene understanding and high-level driving reasoning.
- (Optional) Integrate the developed pipeline with the existing low-level controller provided.
- (Optional) Benchmark the developed framework against existing autonomous driving baselines.
(Tasks can be adapted based on student interests)

Prerequisites:
Strong programming skills in Python.
Experience with deep learning frameworks such as PyTorch or TensorFlow.
Basic knowledge of computer vision, robotics, or autonomous driving systems.
Experience with ROS2, CARLA, reinforcement learning, or sensor fusion.
Motivation to work on modern AI-based autonomous driving systems.

What We Offer:
Access to high-performance computational resources for large-scale autonomous driving simulation and machine learning experiments.
A collaborative research environment within an interdisciplinary autonomous driving and AI research team.
Opportunity to work on cutting-edge Software-Defined Vehicle (SDV) technologies within an EU research project.
Possibility for scientific publications.
Flexible thesis topics depending on the student's background and interests.

If you are interested in the topic, please attach your transcript and CV to the application. We are looking forward to hearing from you.

Kontakt: chengdong.wu@tum.de