Papers
6
Total Citations
543
H-Index
6
About
Mathieu Aubry is a prominent researcher at the intersection of computer vision and robotics, with a particular focus on 3D object understanding, pose estimation, and autonomous robotic manipulation. His work spans several interconnected domains, including grasp planning, visual rearrangement, and robot localization, making significant contributions to the challenge of enabling robots to perceive and interact with the physical world. Aubry's most celebrated contribution is his involvement in the Dexterity Network (Dex-Net 1.0), a cloud-based framework for robust grasp planning that has garnered over 374 citations, underscoring its foundational influence in robotics research. By applying a Multi-Armed Bandit model with correlated rewards, this work demonstrated how cloud robotics infrastructure could be harnessed for scalable, intelligent manipulation planning. Beyond grasping, Aubry has advanced the field of pose estimation through innovative approaches such as RoboPose, which estimates robot joint angles and 6D pose from a single RGB image, and a category-agnostic deep pose estimation method capable of generalizing across arbitrary 3D objects. His research on visually guided rearrangement planning using Monte-Carlo Tree Search further illustrates his commitment to practical, vision-driven robotic autonomy. Collectively, his work reflects a rigorous and imaginative approach to bridging perception and action in intelligent systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3Single-view robot pose and joint angle estimation via render & compare39 citations · 2021
- 4Pose from Shape: Deep Pose Estimation for Arbitrary 3D Objects36 citations · 2019
- 5
- 6