Reinforcement Learning to Mimic Mutually Developments of Infants' Perceptual and Motor Abilities

本研究旨在实现具有更强适用性的强化学习（RL）系统，以减轻设计者在构造RL Agent的状态和动作空间时的负担。首先，我提出了“切换RL模型”来模拟婴儿的运动发展和切换学习模块的方法。此外，我还提出了“发展RL模型”来模拟婴儿的知觉和运动发展，并提出了一种共同构建状态和动作空间的方法。因此，所提出的方法已经实现，以减少设计者的负担，并增加RL的适用性，以真实的世界。

This research aims to realize Reinforcement Learning (RL) system that has more applicability in order to reduce designer's burden when he constructs RL agent's state and action spaces. First, I have proposed "Switching RL model" to mimic an infants' motor development and a method to switch learning modules. Further, I have proposed "Developments RL model" to mimic infants' perceptual and motor developments and a method to co-construct state and action spaces. Therefore, proposed methods have realized to reduce designer's burden and increase an applicability of RL to the real world.