Matthias Zeller

Papers

Total Citations

H-Index

About

Matthias Zeller is a researcher specializing in radar-based perception for autonomous robots and vehicles, with a particular focus on overcoming the limitations of conventional sensing modalities under real-world conditions. His work addresses a critical gap in autonomous navigation: while cameras and LiDAR sensors have driven remarkable advances in scene understanding, they remain vulnerable to adverse weather conditions such as rain, fog, and snow. Zeller's research champions radar as a robust alternative, developing novel deep learning architectures tailored to the unique challenges of sparse and noisy radar point clouds. His most influential contribution, the Gaussian Radar Transformer for semantic segmentation (2022, 35 citations), introduced a transformer-based approach capable of meaningful scene interpretation despite radar's inherent noise characteristics. Building on this foundation, he developed the Radar Instance Transformer (2023, 16 citations), enabling reliable detection and segmentation of moving objects — a capability essential for collision avoidance. His more recent Radar Tracker (2024) extends this pipeline to full instance tracking over time. Together, these works form a cohesive and progressive research agenda that is helping to make autonomous systems more resilient in challenging environments, earning Zeller recognition as an emerging voice in radar-centric robot perception.