Minghai Yao

Papers

Total Citations

H-Index

About

Minghai Yao’s research lies at the intersection of cognitive robotics, reinforcement learning, and computer vision, with a focus on enabling autonomous systems to learn and adapt in dynamic, continuous environments. A central contribution is the development of a Q-learning algorithm integrated with ART2 neural networks, which addresses the “curse of dimensionality” in continuous state-space intelligent systems—a foundational challenge for scalable reinforcement learning. Yao further advanced the field by introducing a visual novelty-driven, internally motivated Q-learning framework for mobile robots, enabling scene learning and recognition that mimics cognitive development, thereby improving learning initiative and adaptability under uncertainty. In object recognition, Yao proposed a SIFT-based algorithm combined with Kalman filtering for robust object matching and identification, directly enhancing robot tracking systems. Although each of these seminal papers has garnered 3 citations, their conceptual novelty—particularly the fusion of neural network architectures with reinforcement learning for cognitive robotics—has laid important groundwork for more adaptive, self-motivated robotic intelligence. Yao’s work is especially notable for pioneering biologically inspired learning paradigms that push beyond traditional supervised methods, offering a blueprint for robots that learn from internal curiosity rather than external labels.