Li Liya, Shi Peng, Ahn Choon Ki
IEEE Trans Cybern. 2022 Jun;52(6):4647-4660. doi: 10.1109/TCYB.2020.3035866. Epub 2022 Jun 16.
For the target-tracking problem, full state of the target may not be available since it may be expensive or impossible to obtain. Thus, the state needs to be reconstructed or estimated only according to measured inputs and outputs. The impossible case that all followers can measure the target directly yields the study of distributed methods, thus reducing the communication and computation resource while resulting in more robustness. This article confronts these problems by addressing a distributed iterative finite impulse response (DIFIR) consensus filter for leader-following systems. A solution to the underlying problem is obtained by involving a distributed measurement model wherein not only the neighbors' estimates are applied but also the directed measurement data are used, and expressed by a computationally efficient iterative algorithm. Applying this DIFIR strategy, it is shown that the leader's estimates by all followers reach H consensus, whose value is the local unbiased estimates of the leader. Then, the result is extended to multiagent systems whose leader has unknown inputs. Incorporating the input estimates, a new DIFIR is proposed. Finally, examples are given to illustrate the consistency and robustness of the developed new design techniques.
对于目标跟踪问题，由于获取目标的完整状态可能成本高昂或根本无法实现，所以目标的完整状态可能不可用。因此，仅需根据测量的输入和输出对状态进行重构或估计。所有跟随者都无法直接测量目标这一不可能出现的情况催生了分布式方法的研究，从而在降低通信和计算资源的同时提高了鲁棒性。本文通过针对领导者 - 跟随者系统提出一种分布式迭代有限脉冲响应（DIFIR）一致性滤波器来解决这些问题。通过引入一种分布式测量模型来获得潜在问题的解决方案，在该模型中，不仅应用邻居的估计值，还使用有向测量数据，并通过一种计算效率高的迭代算法来表示。应用这种DIFIR策略，可以证明所有跟随者对领导者的估计达到H一致性，其值为领导者的局部无偏估计。然后，将该结果扩展到领导者具有未知输入的多智能体系统。结合输入估计，提出了一种新的DIFIR。最后，通过示例说明了所开发的新设计技术的一致性和鲁棒性。
