Papers
36
Total Citations
2,777
H-Index
19
About
Michael Laskey is a robotics researcher whose work sits at the intersection of robot manipulation, deep learning, and imitation learning. He is perhaps best known as a central contributor to the **Dex-Net** project, a landmark series of systems for robust robotic grasping. Dex-Net 1.0 introduced a cloud-based network of 3D object models combined with a Multi-Armed Bandit planning algorithm, while Dex-Net 2.0 — his most influential work with over 1,100 citations — demonstrated that deep neural networks trained on massive synthetic datasets of point clouds and analytic grasp metrics could generalize effectively to real-world objects, dramatically reducing the need for costly physical data collection. Beyond grasping, Laskey has made notable contributions to imitation learning, including DART, a principled noise-injection technique that addresses the compounding error problem in behavior cloning. His research further spans manipulation of deformable objects such as ropes and fabric, surgical robot kinematic calibration, and belief-space motion planning under uncertainty. Across these diverse areas, his work is unified by a commitment to making robot learning more data-efficient and practically deployable. With multiple papers exceeding 75 citations and a body of work that bridges theory and physical experimentation, Laskey has established himself as a significant voice in modern robot learning research.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5Deep Transfer Learning of Pick Points on Fabric for Robot Bed-Making81 citations · 2022
- 6
- 7DART: Noise Injection for Robust Imitation Learning78 citations · 2017
- 8
- 9
- 10