A novel learning-based model-free adaptive control (LMFAC) approach is presented in this article for a class of unknown nonaffine nonlinear discrete-time networked control systems (NCSs) subject to hybrid cyber attacks. The aperiodic denial-of-service (DoS) attacks and persistent deception attacks are assumed to arise in feedback channels, which could result in the absence or authenticity lackness of system signals sent to the controller. With the aid of dynamic linearizaton technology, the equivalent dynamic linearized data models of considered NCSs are first established only based on I/O information instead of the knowledge of mathematical models that are commonly used under the model-based control framework. Then, an LMFAC scheme is designed on the basis of occurred maximum DoS attacks interval to adaptively tune the attenuation coefficient of the input signal for improving system performance during the next DoS attacks interval. Finally, the boundedness of tracking error is rigorously proved through the contraction mapping principle and the effectiveness of the proposed pure data-driven LMFAC method is demonstrated via simulations.

The leaderless or leader-following consensus tracking, along with containment control problems of nonlinear multiagent systems (MASs) using controller-dynamic-linearization-based distributed model-free adaptive control (CDL-DMFAC) method are addressed in this article. By virtue of dynamic linearization (DL) technology, the distributed output and ideal controller of MASs are first converted to the corresponding equivalent DL data models, respectively. Then, a pure data-based CDL-DMFAC scheme is uniformly constructed by employing I/O data regardless of the state space model of MASs. The convergence analysis is rigorously proved by a designed data energy function without using global topology graph information. Furthermore, the control strategy and convergence result are extended to acrlong MIMO MASs. Finally, extensive simulations are performed to verify the validity of theoretical results.

The event-triggered model-free adaptive predictive control (ET-MFAPC) problem for a class of networked nonlinear control systems (NCSs) under deception attacks is addressed in this article. By using dynamic linearization technology, the NCSs are converted to an equivalent data model, and a networked MFAPC scheme with an adjustable input decay rate is constructed to compensate for the network delay. Meanwhile, the attack phenomena existing in feedback channels are modeled by considering both multiplicative and additive deception factors. Then, an ET mechanism without long-time dormancy behavior is proposed to reduce the calculation burden of the controller and save network communication resources. Rigorous convergence analysis for the proposed pure data-driven ET-MFAPC algorithm is given by employing the contraction mapping principle and it shows that the boundedness of tracking error in the mean-square sense can be guaranteed under the presented ET-MFAPC scheme. Finally, extensive simulations are performed to verify the theoretical results.

The consensus tracking and containment control problems of multiple-input and multiple-output (MIMO) nonaffine nonlinear multiagent systems (MASs) are studied in this article under deception attacks using the distributed model-free adaptive control (DMFAC) method. An equivalent dynamic linearized data model of MIMO MASs’s distributed output vector containing deception signals is established using dynamic linearization technology. Then, a fully data-driven DMFAC strategy is designed just using I/O information instead of the knowledge of mathematical model. Furthermore, the boundedness of distributed output vector of MIMO-MASs under deception attacks is proved through the contraction mapping principle without employing global topology graph information. Finally, the validity of the proposed DMFAC scheme is verified through detailed simulations.