Kuniyoshi Yasuo
In contrast to traditional robots, future humanoid robots will not be explicitly programmed to accomplish new tasks in the real world. Instead they will acquire new skills by imitating human behavior. Learning new skills by imitation is a core and fundamental part of human learning, and a great challenge for humanoid robots. This chapter presents mechanisms of imitation learning, which contribute to the emergence of new robot behavior. We first give a detailed description of the general scientific insights underlying the imitation capability of humans. We present a learning model of the mirror neuron system enabling a humanoid robot to imitate first-seen human motions. Then we make further considerations regarding embodiment.
与传统机器人不同，未来的人形机器人不会被明确编程以在现实世界中完成新任务。相反，它们将通过模仿人类行为来获得新技能。通过模仿学习新技能是人类学习的核心和基础部分，也是人形机器人面临的巨大挑战。本章介绍了模仿学习的机制，这些机制有助于新机器人行为的出现。我们首先详细描述了人类模仿能力背后的一般科学见解。我们提出了一种镜像神经元系统的学习模型，使拟人机器人能够模仿首次看到的人类动作。然后我们对具身性进行了进一步的思考。
