基于事件驱动和量化测量的系统辨识

With the fast development of the information technology, it has attracted a lot of attention to study the control theory with event-triggered and quantized observations, but the results on system identification are relatively rare compared to the wide practical needs and application prospects, and then need to be developed and improved urgently. This project will address the system identification with event-triggered and quantized observations, which is an interdisciplinary frontier of the control theory, the communication science and the computer science. For the linear systems, Wiener systems and Hammerstein systems, recursive identification algorithms will be established under event-triggered and quantized observations, and the convergence, convergence rate, asymptotic distribution and asymptotic effectiveness will be obtained after the discussion of identifiability. Importantly, the optimal event-triggered schedule and the optimal quantization scheme will be provided. We will propose the method to derive the minimum amount of data for a given reliability requirement and construct consistent algorithms to estimate it. The complexity problem of system identification will be fully discussed with event-triggered and quantized observations. The adaptive control law will be designed for several kinds of reference signals and the closed performance will be analyzed, including the stability, the robustness and the optimality and so on. The results and methods obtained will be applied to the multi-sensor wireless networks and the multi-agent systems, and the validity of the algorithms will be verified by the simulation and experiment.

随着信息化技术的高速发展，基于事件驱动和量化测量的控制理论研究受到了国内外学者的广泛关注，相对于广泛的实际需求和应用前景，系统辨识方面的成果还相对匮乏、亟待发展和完善。本项目拟研究事件驱动和量化测量下的系统辨识，属于控制理论、通信科学和计算机科学等多领域的前沿交叉课题。具体内容包括：针对线性系统、Wiener系统和Hammerstein系统，构造基于事件驱动和量化测量的递推辨识算法，讨论可辨识性问题，给出收敛性、收敛速度、渐近分布、渐近有效性等，给出最优事件驱动机制、最优阈值、最优量化值的选取方法。建立给定精度下最小数据量的计算方法，构造对其的估计算法，证明收敛性，解决系统辨识的复杂性问题。在多种类型的目标信号下，给出基于事件驱动和量化测量的适应控制律设计方法，分析闭环系统的稳定性、鲁棒性和最优性等。将相关算法应用到多传感器网络和多智能体系统，对其进行仿真和实验，验证有效性和可行性。

本项目是在信息化高度发达和计算机技术迅猛进步下的大环境下提出的，相对于广泛的实际需求，综合考虑事件驱动和数据量化的系统辨识成果还相对匮乏、亟待发展和完善。本项目属于控制理论、通信科学和计算机科学等多领域的前沿交叉课题，在无线传感器网络、数据传输等领域有重要的应用前景。.本项目旨在研究基于事件驱动和量化测量的系统辨识，给出构造未知参数估计算法的方法，并发展相应的控制理论和方法。从线性系统到非线性系统，给出估计算法的收敛性、收敛速度、稳定性和鲁棒性等，并讨论其中的空间和时间复杂性，包括事件机制、输出阈值、量化值、信道噪声、未建模动态等对辨识精度的影响。.在本项目的支持下，还研究了事件驱动机制遇到通信不确定性时的量化系统辨识、面向网络安全的量化系统辨识、基于饱和约束测量的无拖曳控制、复杂动态网络的同步控制等。.本项目资助发表SCI期刊论文16篇，EI论文1篇，会议论文2篇，其中有3篇发表于本领域旗舰期刊IEEE Transactions on Automatic Control和Automatica；资助出版英文著作1部；申请专利5项，其中1项已授权。