Yoshihisa Tsurumine

Nara Institute of Science and Technology

Papers

10

Total Citations

270

H-Index

6

About

Yoshihisa Tsurumine is a leading researcher in deep reinforcement learning (DRL) for robotic manipulation, with a particular focus on deformable objects like cloth. His major contributions center on developing sample-efficient, real-world applicable DRL algorithms that can learn complex control policies directly from high-dimensional raw image inputs, bypassing the need for engineered state representations. His seminal 2018 work on "Deep reinforcement learning with smooth policy update" for robotic cloth manipulation has garnered 179 citations, establishing a foundational approach for stable policy learning in this challenging domain. Tsurumine has pioneered the integration of generative adversarial imitation learning (GAIL) with DRL to overcome the costly burden of manual reward function design, as demonstrated in his 2019 paper (24 citations) and subsequent 2022 work on goal-aware imitation from imperfect demonstrations. He has also advanced sim-to-real transfer through innovative techniques like cyclic policy distillation (2023, 16 citations) and domain randomization, enabling robots to deploy learned policies in the real world with minimal human effort. His notable achievements include implementing DRL on power-efficient FPGA hardware for edge robotics (2021) and extending multi-agent reinforcement learning to cooperative grasping tasks (2025).

Research Focus

Key Achievements

6
H-Index
10
Papers
270
Total Citations
27
Avg Citations/Paper
🏆 Most Cited Paper
Deep reinforcement learning with smooth policy update: Application to robotic cloth manipulation
179 citations · 2018
📈 Most Prolific Year: 2022 (3 Papers)
🤝 Key Collaborators: 14
🏛 Institutions: Nara Institute of Science and Technology

Top Papers

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10

Key Collaborators

Contact & Links

Available for collaboration
Content generated · 15 days ago