Kazuteru Miyazaki
Papers
2
Total Citations
10
H-Index
2
About
Kazuteru Miyazaki is a leading researcher in reinforcement learning, with a focus on developing algorithms that enable efficient learning in complex, real-world robotic systems. His key research areas include online reinforcement learning, penalty-avoiding rational policy making, and continuous-valued input-output systems. Miyazaki’s major contribution is the introduction of **Fixed Mode States** into online reinforcement learning, a method that mitigates the degradation of learning efficiency caused by frequent task failures during long-term trials. This innovation, detailed in his 2012 paper (7 citations), has been applied to biped robot waist trajectory generation, demonstrating practical impact in robotics. He also proposed the **Continuous-Valued Penalty Avoiding Rational Policy Making Algorithm** (3 citations), extending his earlier **Exploitation-oriented Learning (XoL)** framework to handle continuous-valued inputs and outputs, which is critical for real-world applications. By strongly enhancing successful experiences, XoL reduces the number of trial-and-error searches, accelerating learning. Miyazaki’s work bridges theoretical reinforcement learning and practical robotics, offering robust solutions for autonomous systems operating in dynamic environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2