Keita Morisaki

About

Keita Morisaki is a pioneering researcher in robotics and machine learning, whose work focuses on enabling humanoid robots to learn and autonomously generate complex action sequences. His key research areas include recurrent neural networks, reinforcement learning, and robot behavior generation. Morisaki’s major contribution lies in developing a framework that combines Recurrent Neural Networks with Parametric Bias (RNNPB) and reinforcement learning to allow robots to both learn multiple primitive behaviors and automatically sequence them for novel tasks. This approach addresses a fundamental challenge in robotics: moving from pre-programmed actions to adaptive, learned behaviors. His most cited work, "Multiple Action Sequence Learning and Automatic Generation for a Humanoid Robot Using RNNPB and Reinforcement Learning" (2012), has garnered 2 citations and demonstrates how internal parametric bias nodes can encode distinct action primitives, enabling robots to generalize across tasks. While his citation count is modest, Morisaki’s research represents an important step toward more flexible, intelligent humanoid robots capable of learning from experience—a foundation that continues to inspire advancements in autonomous robotics and embodied AI.

Research Focus

Key Achievements

Top Papers

1. 1
   Multiple Action Sequence Learning and Automatic Generation for a Humanoid Robot Using RNNPB and Reinforcement Learning

   2 citations · 2012

Key Collaborators

Contact & Links