Desong Du

Papers

Total Citations

H-Index

About

Desong Du is a researcher focused on advancing autonomous robotics through machine learning and optimization. His primary research areas include deep reinforcement learning for robotic control, task allocation under path constraints, and data-efficient dynamics modeling. Du's most impactful work, "Learning to Control a Free-floating Space Robot using Deep Reinforcement Learning" (2019, 18 citations), introduces a model-free controller that enables free-floating space robots to capture targets without requiring kinematic or dynamic equations—a significant contribution to space robotics where traditional control methods struggle with complex dynamics. In "Homotopic Approach for Robot Allocation Optimization Coupled With Path Constraints" (2019, 9 citations), he addresses a challenging coupling between task allocation and path planning, proposing a homotopy-based method to optimize robot assignments while accounting for path feasibility. His more recent work, "Linear Gaussian Processes for Data-Efficient Robot Dynamics Learning" (2021), explores efficient learning of robot dynamics with limited data, relevant for real-world deployment. Du's contributions demonstrate a commitment to solving practical robotics problems—from space manipulation to multi-robot coordination—by blending control theory, reinforcement learning, and optimization. His research holds promise for autonomous systems operating in unstructured environments.