Papers
16
Total Citations
306
H-Index
8
About
Milad Ramezani is a robotics researcher whose work sits at the intersection of 3D perception, state estimation, and autonomous navigation, with a particular focus on enabling robots to reliably operate in challenging real-world environments. He is perhaps best known for **LoGG3D-Net** (2022, 97 citations), a landmark contribution to LiDAR-based place recognition that advanced how robots localize themselves within pre-built maps using locally guided global descriptors. His **Pronto** framework (2020, 93 citations) demonstrated a robust multi-sensor state estimator for legged robots navigating difficult terrain under demanding conditions, becoming a widely adopted tool in the field. Ramezani has also pushed the boundaries of continual and uncertainty-aware learning for point cloud recognition, addressing the critical challenge of performance degradation in novel, unseen environments. His applied work spans autonomous inspection of industrial and offshore infrastructure using quadruped robots, air-ground collaborative forest localization, and heterogeneous multi-robot teams — most notably contributing to Team CSIRO Data61's top-scoring performance at the prestigious **DARPA Subterranean Challenge**. With over 280 cumulative citations, Ramezani's research meaningfully bridges theoretical advances in 3D perception with practical deployment across field robotics applications.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3InCloud: Incremental Learning for Point Cloud Place Recognition26 citations · 2022
- 4
- 5Air-Ground Collaborative Localisation in Forests Using Lidar Canopy Maps15 citations · 2023
- 6
- 7
- 8
- 9
- 10Uncertainty-Aware Lidar Place Recognition in Novel Environments5 citations · 2023