Eiichi Matsumoto

About

Eiichi Matsumoto is a leading researcher in robotic manipulation and deep reinforcement learning, with a focus on enabling robots to adapt to dynamic, real-world environments. His work bridges the gap between end-to-end learning and practical deployment, most notably through his contributions to the Amazon Robotics Challenge. In his highly cited 2020 paper, "End-to-End Learning of Object Grasp Poses in the Amazon Robotics Challenge," Matsumoto demonstrated how neural networks can directly learn robust grasping strategies from raw sensor data, achieving a 20-citation impact that underscores its relevance to industrial automation. His earlier 2017 work, "Map-based Multi-Policy Reinforcement Learning," introduced a novel framework that allows robots to rapidly adapt to environmental changes or physical damage by switching between pre-trained policies—a critical advancement for mission-critical tasks. With 9 citations, this paper laid the groundwork for more resilient robotic systems. Matsumoto’s research is distinguished by its practical orientation, combining theoretical rigor with real-world testing, making him a key figure in the push toward autonomous, adaptable robots for logistics and beyond.

Research Focus

Key Achievements

Top Papers

1. 1
   End-to-End Learning of Object Grasp Poses in the Amazon Robotics Challenge

   20 citations · 2020
2. 2
   Map-based Multi-Policy Reinforcement Learning: Enhancing Adaptability of Robots by Deep Reinforcement Learning

   9 citations · 2017

Key Collaborators

Contact & Links