Papers
7
Total Citations
87
H-Index
4
About
Mirko Nava is a robotics researcher whose work centers on self-supervised learning, autonomous navigation, and visual perception for unmanned aerial vehicles (UAVs) and robotic systems. His most impactful contribution, "Path Planning With Local Motion Estimations" (45 citations), introduces a novel approach that uses a learned model to predict local motion outcomes from partial knowledge, enabling long-range path planning through self-supervised trajectory data. Nava has pioneered methods for training neural networks with minimal labeled data, as demonstrated in his work on visual localization of quadrotors using noise as self-supervision (15 citations) and his "State-Consistency Loss" technique for spatial perception tasks with partial labels. His research on vision-state fusion (11 citations) advances deep neural networks for autonomous robotics, particularly in challenging scenarios like acrobatic UAV maneuvers. Nava's self-supervised framework for predicting short-range sensor outputs from long-range sensors (2019) and his geometrically interpretable neural perception for visual servoing (2022) further showcase his innovative approach to reducing reliance on expensive labeled data. His recent work on LED state prediction as a pretext task for robot localization (2024) continues this trajectory, making robot perception more accessible and data-efficient.
Research Focus
Key Achievements
Top Papers
- 1Path Planning With Local Motion Estimations45 citations · 2020
- 2
- 3Vision-state Fusion: Improving Deep Neural Networks for Autonomous Robotics11 citations · 2024
- 4
- 5
- 6Visual Servoing with Geometrically Interpretable Neural Perception4 citations · 2022
- 7