网络攻击下不确定信息物理系统的自适应弹性控制

Cyber attacks are key factors threatening the safety of Cyber-Physical Systems (CPS), whose disruption could have a significant impact on public health and lead to large economic losses. Though a number of results have been reported in recent years on secure control of CPS under malicious cyber attacks, most of the results were obtained based on the assumptions that the considered CPS model precisely represents the actual system and is exactly known. Motived by the fact that model uncertainties, regardless of their forms, inevitably exist in practical CPS, the project will study the adaptive resilient control problem of uncertain CPS against typical cyber attacks for guaranteed closed-loop stability and graceful-degradation of CPS operational goals. ..The research objectives are mainly three-folds. .(i).By exploring comparatively the behavior characteristics of multiple malicious cyber attacks including denial-of-service attacks, false data injection attacks and replay attacks, to establish the mathematical formulation of the secure control problem for uncertain CPS under attacks in both single and multiple closed-loop structures. .(ii).By considering the general single-loop and multi-loop CPS models in the presence of parametric uncertainties, to firstly provide rigorous analysis on the effects of typical attacks on representative adaptive control approaches from the perspective of controllability, observability, closed-loop stability and the convergence property of state estimation/regulation/tracking/consensus errors; then provide effective adaptive resilient control strategies in the case of different single attack models; finally provide modified robust adaptive resilient control approaches for the CPS models with external disturbances, unmodeled dynamics and uncertain subsystem interconnections..(iii).Considering the case of switched attack models, to present the unified adaptive attack detection, localization and resilient control design strategy for the general single-loop and multi-loop CPS models in the presence of parametric uncertainties. Besides, to investigate the detection threshold decision approaches based on trade-off between the robustness of external disturbances, unmodeled dynamics, uncertain subsystem interconnections, and the sensitiveness of the attacks. ..Simulation studies of the theoretical results with application to power systems and formation flying UAVs will be conducted.

网络攻击是信息物理系统（Cyber-Physical Systems, CPS）面临的主要安全威胁，可能导致系统瘫痪并对人员、环境和资产造成严重破坏。近年来，自动化领域内的专家学者们在网络攻击下CPS的安全控制研究中已经取得了一些结果。然而，大多数已有结果均假设CPS可以由精确的动态模型刻画。考虑到对实际CPS建模时不可避免的存在各种不确定性，本项目将针对不确定CPS，研究自适应弹性控制方法以保证攻击下闭环系统稳定性和控制性能的得体下降。主要的研究目标是分别在单个和多个闭环CPS模型框架下，（1）建立拒绝服务、虚假数据注入、重放等典型攻击下不确定CPS安全控制问题的数学描述；（2）针对不同的单一攻击模型，提出基于攻击形态重构的自适应弹性控制方法；（3）针对攻击在多个模型间跳变的情况，提出基于攻击检测和定位的切换自适应弹性控制方法。以电力系统和无人机群为应用对象，完成理论结果的仿真验证。

考虑到实际信息物理系统(CPS)不可避免地存在各种不确定性，本项目主要针对含不确定性CPS，研究自适应弹性控制方法以保证攻击下闭环系统的稳定性。项目组在自适应弹性控制、安全检测、自适应量化/事件触发控制、分布式自适应协同控制等方向上取得了多项重要的创新成果，主要包括：（1）针对含未知参数的单闭环二阶非线性CPS，研究了上行链路遭受拒绝服务攻击情况下的自适应弹性控制问题，通过设计自适应观测器规避了攻击下状态反馈信号非连续所带来的设计与分析困难，给出了闭环信号保持全局一致有界的攻击时长强度条件，放宽了已有结果中非线项本质有界的条件；（2）研究了含不确定性高阶内联CPS在欺骗攻击影响双向链路情况下的分散自适应弹性控制问题。引入光滑符号近似函数和新的Nussbaum函数，有效处理了攻击导致的子系统控制系数符号未知、时变且各不相同难题；（3）针对大规模CPS受到的分布式拒绝服务攻击，基于博弈理论提出了一种协同安全检测方法，严格证明了纳什均衡的存在性和唯一性，并提出了纳什均衡迭代学习计算方法；（4）针对含未知参数的单个闭环非线性高阶CPS，研究了上行链路量化通信情况下的状态反馈自适应控制问题，利用无量化情况下虚拟控制的偏导完成了含量化状态情况下的自适应控制器设计，解决了使用非连续状态直接设计反步控制导致的虚拟控制不可导难题。（5）针对含未知参数的一阶非线性多个闭环CPS，在有向拓扑条件下设计了基于事件触发通信机制的完全分布式自适应协同跟踪控制方法，单个子系统设计不需要对于邻居子系统的连续监督，同时考虑子系统内部的未知执行器故障，设计了基于事件触发通信机制的自适应容错协同控制方案。依托本项目支持，已发表SCI论文18篇（含1篇在线发表），EI论文14篇，出版英文专著1部，项目负责人王薇2018年获中国指挥与控制学会科学技术进步奖一等奖（排名一），2020年获批国自然基金优秀青年基金项目。