基于人机交互深层认知信息的多机器人行为协调机制研究

Multi-robot behavior adaptation mechanism based on deep level cognitive information is carried out research in human-robot interaction. According to the correlation analysis between surface level communication information (e.g., speech, gesture, facial expression, and personal identification information) and deep level cognitive information (e.g., communication atmosphere and intention), the key influence factors of communication atmosphere and intention are acquired. And then, to accurately understand the deep level cognitive information, deep level cognitive information understanding models are established based on fuzzy analytic hierarchy process, which include communication atmosphere understanding model based on multi-modal information fusion and intention understanding model by using fuzzy support vector regression. Moreover, for realizing adaptive ability of robots, multi-robot behavior adaptation mechanism is proposed based on communication atmosphere and intention information by using fuzzy inference and multi-agent reinforcement learning. Finally, the applications of deep level cognitive information understanding and multi-robot behavior adaptation mechanism are investigated in the developing network based human-robot interaction system. By studying the project, deep level cognitive information understanding models and multi-robot behavior adaptation mechanisms are put forward for human-robot interaction, and the project would develop some new ideas and schemes for integrating robots into human life, which would be important scientific significance and application value for improving smooth communication between humans and robots.

针对多人对多机器人的人机交互过程，本项目研究一种基于人机交互深层认知信息的多机器人行为协调机制。通过分析表层交流信息（语音、手势、面部表情和个人信息）与深层认知信息（交流氛围和个人意图）的关联性，获取理解交流氛围和个人意图的关键参数；研究基于模糊多层次分析的深层认知信息理解模型，包括基于多模态信息的模糊多层次交流氛围理解模型，以及基于模糊支持向量机的模糊多层次个人意图理解模型，准确理解深层认知信息；结合模糊推理和多智能体强化学习算法，研究基于交流氛围和个人意图的多机器人行为协调机制，实现机器人自适应交流能力；通过建立网络化架构下多人对多机器人的人机交互系统，探讨深层认知信息理解和多机器人行为协调机制的实际应用。本项目的研究将为人机交互过程提供一种理解和适应深层认知信息的有效方法，为机器人融入人类生活和工作中开辟切实可行的新思路和新方案，对于促进人机顺畅交流具有重要科学意义和应用。

面向机器人提高情感能力、实现与人和谐交互的需求，针对人机交互过程产生的表层交流信息和深层认知信息，研究语音、手势、面部表情和个人信息等表层交流信息与交流氛围和个人意图等深层认知信息之间的关联性，建立表层交流信息与深层认知信息之间的关系模型，获取深层认知信息理解模型的关键参数；针对交流氛围和个人意图难以检测的问题，研究逐层实现深层认知信息理解的方法，分别建立基于表层交流信息的交流氛围理解模型和个人意图理解模型；研究基于交流氛围和个人意图的多机器人行为协调机制,实现理解模型和协调机制的仿真分析，开发基于网络的多人对多机器人的人机交互系统，进行深层认知信息理解和适应的实际应用。通过研究，形成基于人机交互深层认知信息的多机器人行为协调机制，实现多机器人的自适应能力，促进人机顺畅交流。