基于人工肌肉的仿人手臂自主控制方法研究

Traditional rigid robot manipulators are usually applied in the static structured and human isolated environment due to their high rigidity and poor safety. Research on artificial muscle-based flexible humanoid arm technology will help to break these limitations and promote robot and human collaboration and coexistence. However, strong hysteresis, creep and other factors existed in artificial muscles make it difficult to accurately describe the dynamic behavior with fixed parameter models, which makes it difficult to make breakthroughs in model-based control methods. Artificial muscles provide passive flexibility to the humanoid arm, while also bringing new challenges to its active compliance control. Moreover, it is a rarely explored and challenging topic to simultaneously estimate the decoupled human motion and stiffness information through surface electromyography and utilize them for humanoid arm control. This project attempts to solve the above problems, and proposes active modeling and control methods for artificial muscles. An active compliance control strategy based on the online-updating model is designed for the passive compliant humanoid arm. Based on motion intention recognition, an autonomous control method for the stiffness of humanoid arm is proposed to reflect humanoid natural flexibility. The proposed methods will provide theoretical basis and technical support for developing high-performance upper limb prostheses and new generation of collaborative robots in the future, which is of great significance for improving the technical level of China's new generation of robots.

传统的刚性机械手臂因其刚性大、安全性差而经常被限定在静态、结构化及与人隔离的环境下工作，而开展基于人工肌肉的柔性仿人手臂技术研究有助于打破这些限制从而促进机器人与人的协作与共融。然而，人工肌肉存在强迟滞、蠕变等因素使得固定参数模型难以精确描述其动态行为，从而造成了基于模型的控制方法难以取得突破性进展；人工肌肉使仿人手臂具有了被动柔顺性，但同时也给主动柔顺控制带来新的挑战；利用表面肌电估计出解耦的运动与刚度信息并将它们融合用于仿人手臂自主控制，是一个鲜有研究且颇具挑战性的课题。本课题尝试解决以上问题，提出人工肌肉主动建模与控制方法，针对被动柔顺性的仿人手臂设计基于在线更新模型的主动柔顺控制策略，基于人体运动意图设计仿人手臂刚度自主控制方法，以体现类人自然柔性。本课题提出的方法为未来研制高性能上肢智能假肢和新一代协作机器人提供理论基础和技术支持，对提升我国新一代机器人的技术水平具有重要意义。

传统的刚性机械手臂因其刚性大、安全性差而经常被限定在静态、结构化及与人隔离的环境下工作，而开展基于人工肌肉的柔性仿人手臂技术研究有助于打破这些限制，使仿人手臂展现出类人的柔性特性，从而促进机器人与人的协作与共融。然而，人工肌肉存在强迟滞、蠕变等因素使得固定参数模型难以精确描述其动态行为，从而造成了基于模型的控制方法难以取得突破性进展；人工肌肉使仿人手臂具有了被动柔顺性，但同时也给主动柔顺控制带来新的挑战；利用表面肌电估计出解耦的运动与刚度信息并将它们应用于仿人手臂自主控制，是一个鲜有研究且颇具挑战性的课题。本课题尝试解决以上问题，提出了人工肌肉柔顺驱动单元的主动建模与控制方法，来抑制人工肌肉特性带来的不确定性，保证驱动系统的控制性能；利用表面肌电信号对连续运动/发力状态进行估计，为仿人手臂柔顺控制提供真实肢体数据；基于协同运动分析进行连续力与刚度的解耦辨识，进而设计仿人手臂的力与刚度独立控制方法，使仿人手臂体现类人自然柔性；最终实现仿人手臂系统的集成与柔顺控制测试，展示了系统整体的类人柔性特性。本课题提出的方法为未来研制高性能上肢智能假肢和新一代协作机器人提供理论基础和技术支持，对提升我国新一代机器人的技术水平具有重要意义。