For multi-agent system (MAS), most of existing iterative learning control (ILC) algorithms consider about the tracking of reference defined over the whole trial interval, while the point-to-point (P2P) task, where the emphasis is placed on the tracking of intermediate time points, has not been explored. Thus, a distributed ILC method is proposed, in which each agent updates the feedforward control input by learning from the experience of itself and its neighbors in previous repeated tasks to achieve the goal of improving performance. In addition, for the sake of reducing the burden of data transmission in MAS, effective data quantization is essential. In this case, the quantitative measurement of the error of the tracking time points is further used in the ILC updating law. In order to accommodate this requirement, a distributed point-to-point iterative learning control (P2PILC) with tracking error quantization for MAS is first proposed in this paper. A necessary and sufficient condition is presented for the asymptotical stability of the proposed algorithm, and simulation results show the effectiveness of it finally.

对于多智能体系统 (MAS)，大多数现有的迭代学习控制 (ILC) 算法都考虑在整个试验间隔内定义的参考跟踪，而点对点 (P2P) 任务则侧重于中间时间点的跟踪，尚未探索。因此，提出了一种分布式 ILC 方法，其中每个代理通过学习自身及其邻居在先前重复任务中的经验来更新前馈控制输入，以达到提高性能的目标。另外，为了减轻MAS中数据传输的负担，有效的数据量化是必不可少的。在这种情况下，在ILC更新律中进一步使用了跟踪时间点误差的定量测量。为了适应这个要求，本文首次提出了一种用于 MAS 的具有跟踪误差量化的分布式点对点迭代学习控制（P2PILC）。给出了算法渐近稳定性的充要条件，仿真结果最终证明了算法的有效性。