Motion Synthesis using Linguistic Creativity towards Robot Intelligence

类人机器人被要求理解与人类相同的语言系统，以便它们融入我们的日常生活。为了实现这一智能机器人，我们在本项目中做了以下研究。为动作和语言之间的联系设计了一个随机模型。人体全身运动被编码到随机模型中，这些模型可以被称为符号。符号与附加在运动上的描述性句子之间的关联由图形随机模型表示。2.用语言理解动作。动作和语言之间的关联框架使机器人能够将观察理解为句子，并合成与给定句子相对应的机器人全身运动。3.相关运动的提取。提取特征动作并对其进行组装，使机器人能够识别各种语言表达的观察结果，并将多个动作组合起来执行复杂的行为。

Humanoid robots are required to understand the same language system as humans so that they are integrated into our daily life. In order to achieve this intelligent robot, we have made following researches in this project.1. Designing a stochastic model for association between motions and language. Human whole body motions are encoded into the stochastic models, which can be referred to as symbols. The association between the symbols and descriptive sentences attached to the motions is represented by graphical stochastic models. 2. Understanding motions using language. The framework of association between motions and language makes it possible for robots to understand observation as sentences and to synthesis robot whole body motion corresponding to given sentences. 3. Extracting Correlated motions. Extracting featured motions and assembling them enable the robots to recognize the observations in various language expressions, and to combine multiple motions to perform complex behaviors.