Unified Benchmarking and Modeling for Automotive Trajectory Prediction

01.04.2026, Abschlussarbeiten, Bachelor- und Masterarbeiten

This master’s thesis focuses on systematically benchmarking major trajectory prediction models in a consistent and structured way, following their development over time. The goal is to build a clear understanding of how different modeling choices affect performance, robustness, and generalization.

Based on this analysis, the project will explore simple and effective improvements or a unified modeling approach that combines the strengths of existing methods.

Key Research Areas and Tasks

• Benchmarking major trajectory prediction models under a unified and reproducible setup
• Analysis of failure cases in trajectory prediction (e.g., collisions, unrealistic trajectories, lack of diversity)
• Development of a clean benchmarking pipeline for fair comparison
• Investigation of modeling improvements or a unified architecture
• Evaluation using common metrics (e.g., ADE, FDE)

Technical Prerequisites (or Motivation to Learn)

• Interest in autonomous driving, machine learning, or robotics
• Basic understanding of deep learning and sequence modeling
• Ability to read and modify research codebases
• Proficiency in Python
• Experience with experiments and data analysis is a plus

Kontakt: Christian.Prehofer@tum.de