Lower-limb exoskeletons often use torque control to achieve easy manipulation of energy flow and ensure human safety. Accuracy of applied torque greatly affects system performance and therefore it is always of interest to be improved. Feed-forward type iterative learning as a compensation term for feedback control was proved effective in torque tracking of these devices with complicated dynamics during human walking, since it is effectively stride-wise integral control. Although it was added merely to benefit average tracking performance over multiple strides, we found that iterative learning after proper gain tuning can help improving real-time torque tracking of lower-limb exoskeletons during human walking. We used theoretical analysis and predicted an optimal gain as the inverse of the passive actuator stiffness. Walking experiments resulted in an optimum 0.9929 $\pm$ 0.3846 times the predicted one, which agreed with the hypothesis. Results of this study provides guidance for the design of torque controller in robotic legged locomotion systems and will help improving the performance of gait assistive robots.

中文翻译：

---

迭代学习增益可优化步态变化下人行走时踝外骨骼的实时扭矩跟踪

下肢外骨骼通常使用扭矩控制来轻松控制能量流并确保人身安全。施加扭矩的精度会极大地影响系统性能，因此始终需要对其进行改进。前馈式迭代学习作为反馈控制的补偿项已被证明可有效地跟踪这些具有复杂运动动力学的设备的转矩，因为它是有效的跨步积分控制。尽管添加它只是为了在多个步幅上提高平均跟踪性能，但我们发现适当的增益调整后的迭代学习可以帮助改善人类行走过程中下肢外骨骼的实时扭矩跟踪。我们使用理论分析并预测了最佳增益作为被动执行器刚度的倒数。步行实验得出的最优值是预测值的0.9929美元\ pm $ 0.3846倍，与假设相符。这项研究的结果为机器人腿式运动系统中的扭矩控制器的设计提供了指导，并将有助于改善步态辅助机器人的性能。