Yuandong Tian

Papers

Total Citations

H-Index

About

Yuandong Tian is a prominent researcher specializing in reinforcement learning, hierarchical learning, and planning-based artificial intelligence systems. His work focuses on developing efficient, generalizable methods for complex sequential decision-making tasks, with particular emphasis on bridging the gap between high-dimensional real-world problems and computationally tractable learning frameworks. Among his notable contributions, Tian has advanced the field of legged locomotion through hierarchical reinforcement learning, demonstrating that fully learned latent action spaces can replace hand-crafted communication interfaces between hierarchical layers, enabling more adaptable and sample-efficient locomotion skills on walking robots. His 2022 work on planning in compact latent action spaces further extended these ideas, addressing a critical scalability challenge by enabling planning-based reinforcement learning to operate effectively in high-dimensional continuous action spaces without prohibitive computational overhead. These contributions reflect Tian's broader mission of making intelligent systems more efficient and generalizable across diverse environments. His research bridges theoretical rigor with practical robotics and AI applications, making his work valuable to both academic researchers and industry practitioners working on autonomous systems, robot control, and deep reinforcement learning. His papers continue to influence ongoing developments in hierarchical and model-based reinforcement learning.