基于柔顺交互认知的机械臂类人阻抗学习自适应控制研究

In order to enhance the performance of compliant interaction control and adaption between robotic manipulator and the unknown environment which is always unstable, the current study will be carried out from the point view of bioinspiration. The adjustment mechanism between CNS (Central Nervous System) and the related muscles of human arms will be studied by modelling and the corresponding experiments for several typical compliant manipulations in dynamic interactive tasks. The coordinating relationship among stiffness, damping and feedforward force for a specific joint will be explored and implemented to the controller design of the manipulator aiming to obtain the human-like compliant behaviors. The following work will be included in this project: Firstly, the activation of the flexor and extensor, contact force as well as trajectory of the arms should be achieved and studied for a typical space dynamic task such as catching a flying ball or plugging a connector. And then the impedance model which can characterize the variation of the physical parameters of the CNS and musculoskeletal system will be founded. Based on this model, the impedance learning strategy can be proposed and optimized. Finally, the effectiveness of the controller will be verified by constructing a testing system for specific tasks for a 6-DOF manipulator. The coordinating mechanisms between the CNS and musculoskeletal system underlying the compliant manipulation are expected to be revealed. In addition, the human inspired impedance learning strategy can be utilized to achieve the desired human-like interaction dynamics. All these studies will provide theoretical evidence and technical support for the improvement of the ability of on-orbit manipulation and extension of the functions of space robots of China.

为提高机械臂对未知、动态、非稳定环境的柔顺交互能力和自适应性，本申请拟借鉴生物灵感，以人体臂部在典型柔顺交互任务中，中枢神经系统和臂部肌肉动作机理开展理论建模和试验研究。研究关节刚度、阻尼以及前馈作用力的协调控制策略并将其移植到机械臂控制器设计中。首先，针对典型柔顺操控过程，研究臂部屈肌、伸肌的活化度、接触力和末端轨迹的变化规律。其次，建立能表征上述物理参数变化规律的中枢神经、骨骼肌肉系统的阻抗模型。进而设计和优化阻抗学习算法。最后，基于项目组的六自由度机械臂，搭建插拔接插件和接球等任务的试验平台，验证和评估上述阻抗学习算法的有效性。项目预期揭示柔顺交互过程中枢神经系统和臂部骨骼、肌肉协调动作机理。基于此设计机械臂自适应阻抗学习控制器，实现所期望的交互动力学特性和类人柔顺操控效果。研究成果将为我国空间机器人在轨操控能力的提升和功能扩展提供理论依据和技术支撑。

故障航天器在轨捕获与维修、轨道碎片离轨清理、在轨加注、空间攻防等精细操作任务已成为制约现代航天技术发展亟待解决的问题之一。面向未知、动态、复杂交互任务的自适应柔顺控制已成为机器人在轨服务能力提升和功能扩展的瓶颈。基于此，本项目围绕上述在轨维护和服务等重大需求，研究了捕获、插拔等典型的面向动态交互过程的机械臂类人阻抗学习自适应柔顺控制策略。建立了一种考虑臂部解剖学结构的柔顺交互力学模型；开发了一种面向典型柔顺操作任务的臂部力/位/肌电信息采集系统；研究了关节刚度、阻尼及前馈作用力的协调控制策略；针对类人臂构型的串联机械臂，设计了一种基于空间周期性的阻抗参数学习算法，并开展了算例仿真验证；以在轨加注的双轴孔插拔操作为例，搭建了主、被动端地面加注平台，对控制策略的有效性开展了实验验证。研究成果可应用于空间机器人精细操作、“人-机”协作机器人等领域。同时，可促进外骨骼机器人设计理论的发展，从而推广到医疗康复领域，服务于人类自身。也可推动生物力学、运动神经生理学等相关学科研究的发展。