Papers
11
Total Citations
142
H-Index
7
About
Erkang Cheng is a robotics and artificial intelligence researcher whose work spans reinforcement learning, autonomous navigation, robot manipulation, and computer vision. His most recognized contribution, "Relay Hindsight Experience Replay" (2023, 38 citations), advances continual reinforcement learning for sequential manipulation tasks under sparse reward conditions — a notoriously difficult challenge in robotic learning. Complementing this, his reward-shaping framework, Dense2Sparse, addresses the critical balance between learning efficiency and effectiveness in uncertain environments. In autonomous driving, Cheng developed a camera-based semantic localization system using HD Maps (2021, 36 citations), offering a cost-effective alternative to expensive sensor suites for precise vehicle positioning. His contributions extend to specialized robotics applications, including pioneering spin estimation and high-speed ball return strategies for table tennis robots, calibration-free vision-based manipulation, stereo matching with pseudo segmentation, and even robotic acupuncture therapy. Notably, his work also touches microrobotics, proposing decoupled magnetic actuation control for untethered microrobots. Collectively accumulating over 130 citations, Cheng's research consistently bridges fundamental machine learning theory with real-world robotic applications, making him a versatile and impactful contributor across multiple frontiers of intelligent robotics.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3A Novel Trajectory-Based Ball Spin Estimation Method for Table Tennis Robot16 citations · 2023
- 4
- 5
- 6
- 7Pseudo Segmentation for Semantic Information-Aware Stereo Matching7 citations · 2022
- 8
- 9
- 10