Y. Tanaka

About

Y. Tanaka is a pioneering researcher in multi-agent reinforcement learning, focusing on accelerating cooperative learning in complex, real-world environments. Their seminal 2002 paper, "Speed up reinforcement learning between two agents with adaptive mimetism" (21 citations), introduced a groundbreaking method that allows agents to dynamically share learning results without homogenizing their behaviors—a critical advance for heterogeneous multi-agent systems. This work demonstrated how selective imitation can dramatically speed up convergence while preserving behavioral diversity. Tanaka further bridged the simulation-to-reality gap in their 2002 study "Propagating learned behaviors from a virtual agent to a physical robot in reinforcement learning" (9 citations), addressing the fundamental challenge of transferring policies from simulated environments to physical robots. By developing techniques to propagate learned behaviors across domains, Tanaka tackled the costly and time-intensive nature of real-world reinforcement learning, where direct physical training is often impractical. Their contributions remain highly influential in robotics and autonomous systems, particularly for applications requiring adaptive, decentralized coordination. Tanaka's research continues to inspire new approaches in transfer learning and multi-agent cooperation, with their work cited as foundational for modern techniques in sim-to-real transfer and collaborative AI.

Research Focus

Key Achievements

Top Papers

1. 1
   Speed up reinforcement learning between two agents with adaptive mimetism

   21 citations · 2002
2. 2
   Propagating learned behaviors from a virtual agent to a physical robot in reinforcement learning

   9 citations · 2002

Key Collaborators

Contact & Links