IEEE Trans Neural Netw Learn Syst. 2017 Mar;28(3):678-689. doi: 10.1109/TNNLS.2015.2511005. Epub 2016 Mar 29.
In this paper, the problem of containment control of networked multiagent systems is considered with special emphasis on finite-time convergence. A distributed neural adaptive control scheme for containment is developed, which, different from the current state of the art, is able to achieve dynamic containment in finite time with sufficient accuracy despite unknown nonaffine dynamics and mismatched uncertainties. Such a finite-time feature, highly desirable in practice, is made possible by the fraction dynamic surface control design technique based on the concept of virtual fraction filter. In the proposed containment protocol, only the local information from the neighbor followers and the local position information from the neighbor leaders are required. Furthermore, since the available information utilized is local and is embedded into the control scheme through fraction power feedback, rather than direct linear or regular nonlinear feedback, the resultant control scheme is truly distributed. In addition, although mismatched uncertainties and external disturbances are involved, only one single generalized neural parameter needs to be updated in the control scheme, making its design and implementation straightforward and inexpensive. The effectiveness of the developed method is also confirmed by numerical simulation.
本文针对网络多智能体系统的牵制控制问题进行了研究，特别关注有限时间收敛性。针对存在未知非仿射动态和不匹配不确定性的情况，提出了一种分布式神经网络自适应牵制控制方案。与当前的研究现状不同，该方案能够在有限时间内以足够的精度实现动态牵制，具有很强的实用性。这种有限时间特性是通过基于虚拟分数滤波器概念的分数动态面控制设计技术实现的。在提出的牵制协议中，只需要邻居跟随者的局部信息和邻居领导者的局部位置信息。此外，由于所利用的信息是局部的，并通过分数幂反馈嵌入到控制方案中，而不是直接的线性或正则非线性反馈，因此所得到的控制方案是真正分布式的。此外，尽管存在不匹配的不确定性和外部干扰，但控制方案中只需要更新一个单一的广义神经参数，这使得其设计和实现简单且经济。数值仿真也验证了所提出方法的有效性。
