Behaviors Acquisition for Multiple Tasks by Integration of Imitation Learning and Reinforcement Learning

点击翻译按钮获取中文摘要

In this research, we realized the effective acquisition of multiple behaviors for real robots by integration of imitation learning and reinforcement learning. First of all, we realized the behavior learning for single task such as 1) integration of imitation learning and behavior learning for the measuring worm robots and 2) integration of CPG and reinforcement learning for the four-legged robots. Moreover, we realized the behavior learning for multiple tasks such as 3) imitation learning of multiple behaviors for the two-armed gating robots and 4) effective acquisition of multiple behaviors by integration of CPG and reinforcement learning for the four-legged robots.