Direkt zum Inhalt springen

Master’s Thesis – Real-Time Trustworthiness Estimation for V2X Participants Using Multi-Modal Confidence Sources (in cooperation with BMW)

08.08.2025, Abschlussarbeiten, Bachelor- und Masterarbeiten

The Chair of Robotics, Artificial Intelligence, and Real-Time Systems offers a Master’s thesis in cooperation with BMW. The topic is real-time trust estimation in V2X using motion data, environment info, detection confidence, and latency. A dynamic trust score is computed in-vehicle and at intersections to guide cooperative perception and localization.

Motivation & Relevance

In V2X systems, not all transmitted data is equally reliable. Environmental factors, sensor fusion and object detection confidence, as well as communication latency all impact the perceived trustworthiness of a vehicle's message. A robust trustworthiness estimation mechanism allows infrastructure and vehicles to make smarter decisions about which participants to prioritize in tasks such as localization or perception fusion—especially under uncertainty or potential misbehavior. This is crucial for safety-critical applications and the integrity of cooperative driving systems.

Project Description

You will design a real-time system capable of:

  • Gathering relevant confidence sources (e.g., movement consistency, VLM weather info, sensor fusion confidence, object detection certainty, communication latency)
  • Creating a dynamic trust score model executable both in-vehicle and on the infrastructure side (e.g., roadside units)
  • Merging local and external trust scores to compute a global confidence level per vehicle
  • Implementing a real-time preference mechanism for perception and localization based on these scores
  • Exploring machine learning or reinforcement learning to adapt scoring under changing conditions

Your Tasks

  • Develop a trust score model using multi-modal input data
  • Implement a prototype for real-time estimation (Python, C++, or embedded target)
  • Integrate reinforcement learning or adaptive logic for dynamic adjustment of scoring parameters
  • Simulate and evaluate the system under different traffic/weather scenarios (e.g. using CARLA)
  • Analyze latency impact and ensure real-time viability of the system

Profile

  • Master’s student in Computer Science, Robotics, Electrical or Communication Engineering
  • Proficiency in Python or C++
  • Basic understanding of real-time or embedded systems
  • Interest in V2X communication, trust modeling, and intelligent systems
  • Understanding of perception pipelines (sensor fusion, object detection) is beneficial
  • Experience with ML/RL frameworks (e.g., PyTorch, TensorFlow) is a plus

What You Will Gain

  • Hands-on experience in real-time trust estimation for V2X applications
  • Exposure to multi-modal data integration and real-time decision systems
  • Work on a project with an industrial partner focused on future-oriented V2X safety and perception architectures
  • Insights into reinforcement learning for dynamic decision-making in embedded environments

How to Apply

Please send your CV and a transcription of your grades within your application.

Kontakt: erik-leo.hass@tum.de