Papers
11
Total Citations
88
H-Index
6
About
Pu Feng is an emerging researcher whose work sits at the dynamic intersection of multi-agent reinforcement learning (MARL), multi-robot cooperation, and robust autonomous systems. With a growing body of highly cited publications, Feng has established a distinctive research identity centered on improving the sample efficiency, generalization, and robustness of cooperative MARL frameworks. Among Feng's most notable contributions is a series of works exploiting symmetry as an inductive bias in multi-agent settings — spanning perfect symmetry, partial symmetry, and hierarchical symmetry — to dramatically reduce data requirements and improve physical consistency in learned policies. His hierarchical consensus-based MARL framework, which bridges the gap between centralized training and decentralized execution, has garnered 18 citations since its 2024 publication, reflecting strong community interest. Feng has also advanced collision avoidance through continuous-transition reinforcement learning and explored adversarial robustness by studying minority influence attacks on cooperative agents. More recently, his Lyapunov-informed MARL approach introduces control-theoretic stability priors to accelerate training convergence. Collectively accumulating over 85 citations across just a few years, Feng's research offers valuable tools for deploying reliable multi-robot systems in real-world environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2Leveraging Partial Symmetry for Multi-Agent Reinforcement Learning15 citations · 2024
- 3ESP: Exploiting Symmetry Prior for Multi-Agent Reinforcement Learning10 citations · 2023
- 4
- 5
- 6
- 7
- 8
- 9
- 10Exploiting Hierarchical Symmetry in Multi-Agent Reinforcement Learning3 citations · 2024