Chengzhong Ma
Papers
2
Total Citations
10
H-Index
2
About
Chengzhong Ma is a researcher advancing the frontiers of robot manipulation through reinforcement learning and hierarchical planning. His work primarily addresses two critical challenges in robotics: enabling safe, data-efficient learning and improving multi-object manipulation in cluttered environments. Ma’s most cited paper, “Improving Offline Reinforcement Learning With in-Sample Advantage Regularization for Robot Manipulation” (2024, 6 citations), tackles the core problem of offline RL—learning effective policies from fixed datasets without risky real-world exploration. By introducing in-sample advantage regularization, his method enhances both learning efficiency and safety, a vital step for deploying robots in real-world settings. In “Prioritized Planning for Target-Oriented Manipulation via Hierarchical Stacking Relationship Prediction” (2023, 4 citations), Ma addresses the complexity of grasping multiple targets by proposing a novel approach to predict and leverage hierarchical stacking relationships between objects. This work enables robots to plan safer and more efficient manipulation sequences in scenes where objects are densely stacked. With a growing citation record and a focus on practical, safety-critical robotics, Ma’s contributions are paving the way for more intelligent and autonomous robotic systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2