Abstract Collaborative tracking control involves two or more subsystems working together to perform a global objective, and is increasingly used within a diverse range of applications. Decentralised iterative learning control schemes have demonstrated highly accurate collaborative tracking by using past experience gained over repeated attempts at the task. However they impose highly restrictive constraints on the system dynamics, and their reliance on inverse dynamics has degraded their robustness to model uncertainty. This paper proposes the first general decentralised iterative learning framework to address this problem, thereby enabling a wide range of existing iterative learning control methodologies to be applied in a decentralised manner to collaborative subsystems. This framework is illustrated through the derivation of a variety of new decentralised iterative learning control algorithms which balance collaborative tracking performance with optimisation of a general objective function. The framework is illustrated by application to wearable stroke rehabilitation technology in which each subsystem is a muscle artificially activated by electrical stimulation. These verify the framework’s simplified design and reduced hardware and communication overheads.

中文翻译：

---

一种用于协同跟踪的去中心化迭代学习控制框架

摘要 协同跟踪控制涉及两个或多个子系统协同工作以执行全局目标，并且越来越多地用于各种应用中。分散式迭代学习控制方案通过使用在重复尝试任务中获得的过去经验证明了高度准确的协作跟踪。然而，它们对系统动力学施加了高度限制性的约束，并且它们对逆动力学的依赖降低了它们对模型不确定性的鲁棒性。本文提出了第一个通用的分散式迭代学习框架来解决这个问题，从而使广泛的现有迭代学习控制方法能够以分散的方式应用于协作子系统。该框架通过各种新的分散式迭代学习控制算法的推导来说明，这些算法平衡了协作跟踪性能与一般目标函数的优化。该框架通过应用于可穿戴中风康复技术来说明，其中每个子系统都是由电刺激人工激活的肌肉。这些验证了框架的简化设计并减少了硬件和通信开销。