Master's Thesis - V2X: Offline Generation of ETSI-Compliant Collective Perception Messages (CPMs) from Real-World and Synthetic Datasets (cooperation with BMW)

26.11.2025, Diplomarbeiten, Bachelor- und Masterarbeiten

The Chair of Robotics, Artificial Intelligence, and Real-Time Systems offers a Master’s thesis focused on generating ETSI-compliant Collective Perception Messages (CPMs) from offline V2X datasets. The goal is to design a modular pipeline that converts annotated objects into CPM structures, derives dynamic attributes, models uncertainties, and provides a reproducible CPM-ready data framework to support cooperative perception, fusion, and communication research.

Motivation & Relevance

ETSI Collective Perception Messages (CPMs) enable vehicles and infrastructure to extend their field of perception beyond the limits of on-board sensors. While ETSI defines a clear standard for CPMs, most real-world and synthetic V2X datasets do not provide any standardized mechanisms to generate CPMs offline. As a result, researchers may create custom tooling with inconsistent assumptions about uncertainty modeling, object aging, dynamic quantities, or sensor attribution. This fragmentation limits reproducibility, hinders meaningful benchmarking, and makes it difficult to compare cooperative perception systems fairly. A unified, modular pipeline that generates ETSI-compliant CPMs from real V2X datasets would fill a major gap in current research - enabling consistent evaluation of cooperative perception, multi-sensor fusion, and communication efficiency.

Project Description

In this thesis, you will design and implement a scalable offline CPM generation framework that processes existing V2X datasets (e.g., DAIR-V2X-C, V2X-Seq, TUMTraf-V2X). In addition, you will include an existing CARLA pipeline, which is aiding your offline generation.

Your system will:

• implement a modular architecture (dataset loader → converter → CPM encoder)
• compute missing dynamic quantities (e.g., velocity, acceleration) based on bounding boxes and timestamps
• design uncertainty and object aging models that match ETSI CPM requirements
• model sensor contribution and mapping logic
• support both real-world and simulation-based workflows
• evaluate message completeness, consistency, and reproducibility

The final result will be a reusable and well-documented pipeline that supports future research in cooperative perception and V2X communication.

Your Tasks

• Design the CPM generation architecture
• Derive motion quantities from object positions and timestamps
• Develop uncertainty models and object aging mechanisms
• Implement dataset-independent workflows
• Evaluate CPM quality and reproducibility across datasets

Your Profile

• Master’s student in Computer Science, Electrical Engineering, Robotics, or a related field
• Proficiency in Python
• Interest in V2X systems and cooperative perception
• Experience with dataset processing or perception pipelines is beneficial

WHAT You WILL GAIN 

• Hands-on experience with ETSI CPM structures and message generation
• Expertise in uncertainty modeling and motion reconstruction
• Insight into V2X communication, cooperative perception, and dataset harmonization
• Practical skills for research and industrial development in automated and connected driving
• Chance to contribute to a real-world-relevant project (with BMW)
  

Kontakt: erik-leo.hass@tum.de