Papers
11
Total Citations
359
H-Index
6
About
Kei Kase is a leading researcher in humanoid robotics and deep learning for robot manipulation, whose work focuses on enabling robots to perform complex, multi-step tasks with human-like dexterity and adaptability. Kase’s most influential contribution is the development of a machine-learning-based humanoid robot capable of working as a production line worker, as detailed in their highly cited 2016 paper (235 citations), which introduced a practical method for repeatable folding tasks. This foundational work has been extended through pioneering research on transferable task execution from raw visual input using deep planning domain learning (40 citations), allowing robots to solve novel problems beyond their training data. Kase has also made significant strides in compound task generation, demonstrating how humanoid robots can execute multiple discrete tasks in sequence using deep neural networks (31 citations). More recently, their innovative work on material classification using active temperature-controllable robotic grippers (9 citations) and active heat flow sensing (6 citations) has opened new frontiers in multimodal object recognition. Kase’s research, which bridges deep learning, cognitive robotics, and sensorimotor control, has garnered over 350 citations, establishing them as a key figure in advancing practical, generalizable robot manipulation for real-world applications.
Research Focus
Key Achievements
Top Papers
- 1Repeatable Folding Task by Humanoid Robot Worker Using Deep Learning235 citations · 2016
- 2
- 3
- 4
- 5
- 6
- 7Active Heat Flow Sensing for Robust Material Identification6 citations · 2023
- 8
- 9Leveraging Motor Babbling for Efficient Robot Learning5 citations · 2021
- 10Use of Action Label in Deep Predictive Learning for Robot Manipulation4 citations · 2022