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

7
H-Index
11
Papers
142
Total Citations
13
Avg Citations/Paper
🏆 Most Cited Paper
Relay Hindsight Experience Replay: Self-guided continual reinforcement learning for sequential object manipulation tasks with sparse rewards
38 citations · 2023
📈 Most Prolific Year: 2022 (5 Papers)
🤝 Key Collaborators: 33
🏛 Institutions: University of Science and Technology of China, Zhengzhou University, Chinese Academy of Sciences, Institute of Intelligent Machines

Top Papers

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10

Key Collaborators

Contact & Links

Available for collaboration
Content generated · 13 days ago