Research On Adaptive Security Control Of Switched Nonlinear Systems Under Network Attack

In recent years,network communication has been widely applied for information exchange among system components(such as sensors,controllers,actuators,etc.)due to its advantages in real-time data transmission,remote monitoring,highly flexible system integration,and scalability,thereby establishing a connection between physical systems and cyberspace.While offering numerous conveniences,network communication inevitably poses more severe challenges to the security and reliability of modern industrial systems.However,current research on control system safety primarily focuses on non-switched linear or nonlinear systems with limited reports addressing switched nonlinear systems that can characterize complex practical industrial applications like robots,unmanned aerial vehicles(UAVs),and manipulators.On one hand,the introduction of network transmission makes the study of switched systems more complex.In addition to addressing issues arising from the interaction between continuous dynamics and discrete dynamics,coupled with switching signals and control laws,such as system stabilization and asynchronous control,researchers will also face challenges induced by networkrelated constraints,energy limitations,transmission disruptions,and network attacks.On the other hand,in practical engineering,he co-design of control laws and switching signals for uncertain switched systems is a challenging problem due to the need for uncertainty modeling and consideration of various constraints(such as saturation constraints and prescribed performance constraints).Hence,it is an important and practical research topic to explore adaptive security control problems in switched nonlinear systems within a network transmission environment.The thesis uses neural network or fuzzy logic system approximation method,Backstepping technique and dwell time method to study the adaptive security control problem of switched nonlinear system under network transmission when it is under Denial of Service(DoS)attacks and deception attacks.The main contents are as follows:1.The event-triggered security problem under DoS attacks is investigated for a class of switched nonlinear systems with strict feedback uncertainty.First,a set of switching adaptive neural network observers is designed to estimate the unmeasured state and compensate for the continuously blocked output signal caused by energy-limited attack behavior in the sensorcontroller network.The output of the designed observer is the real system output when the attack sleeps,and it provides an estimate during active attacks.Furthermore,considering limited network transmission resources,an adaptive neural network event-triggered controller is constructed using the designed observer and Backstepping technology,which together with the common Lyapunov function stability theory,achieves the semi-global uniform final boundedness of the closed-loop system under arbitrary switching signals.2.The adaptive security control problem is considered for a class of switched nonlinear systems with unmodeled dynamics under DoS attacks and restricted switching signals.An adaptive neural network switching observer is designed to compensate for system information blocked by the attack in the modeled part.Assumptions are made for the unmodeled part,and an estimation of the unknown nonlinearity is obtained using inequality reduction and neural network techniques.Multiple Lyapunov candidate functions are selected to construct both virtual controllers and adaptive neural network controllers through the Backstepping technique.Additionally,the average dwell time switching signal related to the attack duration and frequency is designed to address the deep coupling problem between the switching behavior of the considered system and class-switching caused by DoS attacks,while ensuring uniform boundedness of all signals in the closed-loop system with respect to the designed dwell time switching signal.3.The asymptotic tracking control based on preset performance is addressed for a class of uncertain switched nonlinear systems subject to input saturation constraints.An auxiliary system is established based on the input saturation errors,and a new preset performance function with modified signals is designed to characterize the potential relationship between input saturation and performance constraints.The designed performance function adaptively expands or restores the user-specified performance function boundary depending on whether saturation occurs.By introducing the error transfer function,an obstacle Lyapunov function is constructed,and Backstepping technology with command filter is employed to construct virtual controllers and an adaptive neural network event-triggered controller using the first-order filter and compensators.The control approach achieves:1)semi-global uniform final boundedness of the closed-loop system under arbitrary switching signals;2)asymptotic convergence of tracking error to zero and fixed-time predefined performance control.Under the proposed prescribed performance function,we further consider deception attacks in sensor-controller networks for such switched nonlinear systems.By utilizing Nussbaum functions and neural network approximation techniques,the unknown control direction problem caused by deception attacks is addressed,removing the assumption that attack gain must be greater than zero.Finally,based on neural network approximation methods and Backstepping technology,we propose adaptive virtual controllers and input saturation controllers,achieving fixed-time prescribed performance control of the closed-loop systems.4.The output feedback security control based on prescribed performance under deception attacks is considered for a class of uncertain switched nonlinear systems with unmeasured states and input saturation.First,a set of adaptive fuzzy-switching observers is designed using the system output information to estimate the unmeasurable state.Subsequently,an auxiliary system is constructed to design a prescribed performance function that captures the relationship between input saturation and performance constraints.When no saturation occurs,the auxiliary system can restore the designed performance function to the userspecified boundary within a finite time.Based on the constructed switching observer and prescribed performance function,virtual controllers and adaptive fuzzy controllers are established using the Backstepping technique.Moreover,employing multi-Lyapunov function stability theory and mode-dependent average dwell time method,we propose a less conservative mode-dependent average dwell time switching signal compared to existing literature.By combining the switching signal with the adaptive fuzzy security controller,the semi-global uniform final boundedness of the closed-loop system under such switching signals is achieved,while ensuring that the tracking trajectory enters the boundary of the proposed prescribed performance function at a fixed time.