Minghao Yu

Northeastern University, Chinese University of Hong Kong

Papers

Total Citations

H-Index

About

Minghao Yu is a robotics researcher whose work focuses on bridging the gap between human dexterity and robotic intelligence, particularly in the domains of robotic manipulation, humanoid skill transfer, and long-horizon task planning. His key contributions include developing a light-weight CNN model for real-time robotic grasp detection (2020, 8 citations), which significantly improves the efficiency of industrial applications like assembly and sorting. Yu also introduced HOTU, a groundbreaking framework for cross-embodiment behavior-skill transfer between humans and humanoid robots using decomposed adversarial learning from demonstration (2025, 5 citations), addressing the challenge of learning complex loco-manipulation tasks in high-degree-of-freedom systems. More recently, he has advanced the field of instruction-following long-horizon manipulation by integrating LLM-empowered symbolic planners for dual-arm mobile robots (2024, 2 citations), enabling versatile, language-guided task execution in real-world environments. His work is notable for its practical emphasis on real-time performance and scalability, with applications ranging from industrial automation to assistive robotics. Yu’s research represents a significant step toward creating embodied intelligent agents capable of performing human-level tasks autonomously.