Keisuke Azetsu
Papers
2
Total Citations
5
H-Index
2
About
Keisuke Azetsu is a researcher advancing the frontiers of multi-robot coordination and cooperative transportation through deep reinforcement learning. His work focuses on enabling robot teams to autonomously execute complex formation changes during transport tasks—a critical capability for real-world logistics and automation. Azetsu’s major contributions include developing accelerated learning frameworks that address the high sample complexity of traditional reinforcement learning. In his 2024 paper, he introduced SDPA-MAPPO (Scaled Dot Product Attention-Multi-Agent Proximal Policy Optimization), a transfer learning method that significantly speeds up formation change adaptation for cooperative transport, garnering early attention with 3 citations. His earlier 2022 work on DeepDyna-Q proposed a model-based approach to fast learning in unknown environments, achieving 2 citations. These innovations tackle the fundamental challenge of making multi-robot systems learn efficiently and adapt quickly, moving beyond popular methods like MADDPG. Azetsu’s research is pivotal for students and engineers seeking to deploy scalable, intelligent robot teams in dynamic settings, demonstrating how attention mechanisms and model-based learning can unlock practical, real-time cooperative behaviors.
Research Focus
Key Achievements
Top Papers
- 1
- 2