Jiafei Lyu

About

Jiafei Lyu’s research lies at the intersection of reinforcement learning (RL) and combinatorial optimization, with a particular focus on multi-goal RL and multi-entity task allocation. Lyu has made significant contributions to overcoming two persistent challenges in multi-goal RL: sparse rewards and sample inefficiency. Their work on bias-reduced multi-step Hindsight Experience Replay (HER) introduces novel goal relabeling strategies that improve learning efficiency in planning and robot manipulation tasks, directly addressing the limitations of standard HER approaches. In the domain of task allocation, Lyu developed a two-stage RL-based framework that moves beyond static assumptions, enabling dynamic and scalable allocation of entities to tasks in complex, real-world scenarios such as multi-robot cooperation and resource scheduling. With a growing citation record—including 19 citations for their 2024 task allocation paper—Lyu’s research is gaining recognition for its practical impact. Their work not only advances algorithmic foundations but also offers deployable solutions for autonomous systems, making it highly relevant for students and researchers interested in sample-efficient RL and multi-agent coordination.

Research Focus

Key Achievements

Top Papers

1. 1
   A two-stage reinforcement learning-based approach for multi-entity task allocation

   19 citations · 2024
2. 2
   Multi-Step Hindsight Experience Replay with Bias Reduction for Efficient Multi-Goal Reinforcement Learning

   3 citations · 2023
3. 3
   Bias-reduced Multi-step Hindsight Experience Replay for Efficient Multi-goal Reinforcement Learning

   2 citations · 2021
4. 4
   A Two-stage Reinforcement Learning-based Approach for Multi-entity Task Allocation

   2 citations · 2024

Key Collaborators

Contact & Links