Pusong Li

Papers

Total Citations

H-Index

About

Pusong Li is a robotics researcher whose work sits at the intersection of grasp planning, cloud robotics, and learning from demonstration. His research addresses fundamental challenges in enabling robots to reliably manipulate a wide variety of objects, particularly in unstructured environments. Li’s most notable contribution is the introduction of “adversarial grasp objects” (2019, 30 citations), a concept inspired by adversarial examples in computer vision that identifies objects physically similar to target items but significantly more difficult for learned grasp planners to handle—a critical insight for improving the robustness of deep learning-based grasping systems. He also developed Dex-Net as a Service (DNaaS) (2018, 25 citations), a pioneering cloud-based grasp planning system that provides a graphical interface for parallel-jaw grippers, reducing infrastructure overhead and democratizing access to advanced grasp planning algorithms. In earlier work (2017, 23 citations), Li leveraged the da Vinci Research Kit (dVRK) teleoperation to facilitate deep learning of automation tasks for industrial robots, demonstrating how human demonstration data can be effectively transferred to robot kinematics for precise bilateral manipulation. Together, these contributions have helped shape the trajectory of robust, data-driven robotic manipulation.