Knacks and Focuses of Whole Body Movement --- Humanoid Science

在本研究中，我们将“诀窍”和“焦点”视为行为动力学和信息结构上的“稀疏控制点”和“控制信息的稀疏提取点”，以实验澄清、理论化和构建应用为目的。我们得到了以下结果：澄清人体全身行为技能的“诀窍”和“重点”：我们通过整合我们的工作，包括人体运动的测量和分析，以及动态模型的模拟和分析，揭示了运动任务的动态状态轨迹的整体结构与表现成功条件之间的关系（澄清“诀窍”）。我们通过定量实验发现，人类在观察他人的动作并判断成功或失败时，会关注与控制点相似的点。（澄清“重点”）用成人大小的人形机器人实现全身动态运动：我们成功完成了世界上第一个成人大小的人形机器人实现动态上下运动的实验。通过分析数百个真实机器人实验的数据，我们揭示了一个区分性能（成功或失败）的特征。在实现上述目标的过程中，我们还为高动态机器人系统开发了原始组件技术，包括能够产生非常高的冲激力的驱动系统，非常快速的控制系统，以及用于检测与环境接触状态的分布式触摸传感器。“小”与“焦点”理论及自动提取方法：基于上述结果1。和2。研究了“小窍门”和“重点”的集成理论，构建了一种自动提取方法。从全局动力学的角度来看，我们证明了“Knacks”和“焦点”指的是运动轨迹上的同一点。这一点将是动作模仿学习以及最小干扰但有效的任务辅助的关键。基于任务状态空间的可达性概念，提出了一种自动生成捕获“Knack”的运动控制策略的方法。仿真实验验证了该方法的有效性。总之，这个项目非常成功，取得了世界领先的成果，包括第一次科学地阐明“Knacks”和第一次看到机器人的运动，在学术界得到了很高的评价，在大众媒体上也出现了。少

In this research, we have regarded "Knacks" and "Focuses" as "Sparse control points" and "Sparse extracted points of control information" on dynamics and information structure of behavior, and have aimed at the experimental clarification, theorization, and construction of application. We have obtained the following result.1.Clarification of "Knacks" and "Focuses" on Skills of Human Whole Body Behaviors : We revealed a relationship between a global structure of dynamical state trajectory of a movement task and the success condition of performance, applied to dynamic roll-and-rise motion, by integration of our works including measurements and analyses of the human movement, and simulations and analyses of the dynamical model (Clarification of "Knacks"). And we revealed by our quantitative experiment, that humans focus on similar points to the control point when they observe other person's motion and judge success or failure of it. (Clarification of "Focuses").2.Realizing Whole Body Dynam … More ic Motion by an Adult-size Humanoid Robot : We succeeded the world's first experiment in which an adult-size humanoid robot realized dynamic roll-and-rise motion. And we revealed a feature which discriminates the performance (success or failure), by analyzing the data from hundreds of real robot experiments. In realizing the above, we also developed original component technology for the highly dynamic robot system, including a drive system capable of very high impulsive force, a very fast control system, and a distributed touch sensor for detecting contact states with the environment.3.The Theory of "Knacks" and "Focuses" and The Auto Extracting Method : Based on results above 1. and 2., we investigated an integration theory of "Knacks" and "Focuses", and constructed an auto extracting method. From view point of global dynamics, we showed that a "Knacks" and "Focuses" refer to the same point on a motion trajectory. This point will be the key to imitation learning of motion as well as minimally obtrusive but effective task assistance. We proposed a method which generates a motion control strategy capturing "Knack" automatically, based on the notion of reachability in the task state space. The effectiveness of the method was confirmed by simulation experiments.To summarize, this project was quite successful with world-leading results including the first scientific clarification of "Knacks" and the first-seen robot motion, manifested by high estimation in the academia and appearance in mass-media. Less