Kuniyoshi Yasuo, Yorozu Yasuaki, Suzuki Shinsuke, Sangawa Shinji, Ohmura Yoshiyuki, Terada Koji, Nagakubo Akihiko
Department of Mechano-Informatics, School of Information Science and Technology, The University of Tokyo, 7-3-1, Hongo, Bunkyo-ku, Tokyo, 113-8656, Japan.
Prog Brain Res. 2007;164:425-45. doi: 10.1016/S0079-6123(07)64023-0.
A constructivist approach to cognition assumes the minimal and the simplest set of initial principles or mechanisms, embeds them in realistic circumstances, and lets the entire system evolve under close observation. This paper presents a line of research along this approach trying to connect embodiment to social cognition. First, we show that a mere physical body, when driven toward some task goal, provides a clear information structure, for action execution and perception. As a mechanism of autonomous exploration of such structure, "embodiment as a coupled chaotic field" is proposed, with experiments showing emergent and adaptive behavior. Scaling up the principles, a simulation of the fetal/neonatal motor development is presented. The musculo-skeletal system, basic nervous system, and the uterus environment are simulated. The neural-body dynamics exhibit spontaneous exploration of a variety of motor patterns. Lastly, a robotic experiment is presented to show that visual-motor self-learning can lead to neonatal imitation.
一种建构主义的认知方法假定了最少量且最简单的初始原则或机制,将它们置于现实情境中,并让整个系统在密切观察下发展。本文展示了沿着这种方法的一系列研究,试图将具身性与社会认知联系起来。首先,我们表明,仅仅一个物理身体,当被驱动朝着某个任务目标行动时,会为行动执行和感知提供清晰的信息结构。作为对这种结构进行自主探索的一种机制,我们提出了“具身性作为耦合混沌场”,实验表明了涌现和适应性行为。将这些原则进行扩展,我们展示了一个胎儿/新生儿运动发育的模拟。对肌肉骨骼系统、基本神经系统和子宫环境进行了模拟。神经 - 身体动力学展现出对各种运动模式的自发探索。最后,展示了一个机器人实验,以表明视觉 - 运动自我学习可以导致新生儿模仿。