Shuhao Liao

Papers

Total Citations

H-Index

About

Shuhao Liao is pioneering the intersection of multi-agent reinforcement learning (MARL) and embodied intelligence for aerial swarms. His research focuses on two critical frontiers: leveraging structural inductive biases to improve learning efficiency, and bridging the sim-to-real gap for large-scale robotic systems. In his highly cited 2024 work, Liao introduced the concept of leveraging *partial* symmetry—rather than perfect symmetry—as an inductive bias in MARL, demonstrating significant improvements in generalization, data efficiency, and physical consistency for cooperative tasks. This breakthrough addresses a fundamental limitation of prior methods that assumed rigid, perfect symmetry in multi-agent domains. Complementing this theoretical advance, Liao developed Air-M (2023), a visual reality platform designed specifically for training and deploying reinforcement learning policies on large-scale aerial unmanned systems. Air-M tackles the dual challenges of high sample complexity and difficult sim-to-real transfer, providing a critical infrastructure for advancing autonomous drone swarms. With 15 citations on his symmetry paper and 4 on Air-M, Liao’s work is rapidly gaining recognition for its practical impact on scalable, real-world multi-agent systems.