Learning‐based robust control methodologies under information constraints

Karimi, Hamid Reza; Ning, Wang; Man, Zhihong

doi:10.1002/rnc.5973

Cited by 3 publications

(2 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [24], the event-based security control issues in discrete-time NCSs were addressed to reduce the influences of deception attacks and DoS attacks. Different from the above results, this paper applies a game-theoretic approach to design a secure tracking control scheme, which is capable of simultaneously resisting the influence of DoS attacks and external disturbance [25][26][27]. However, as far as we know, limited works can be found on the topic of secure tracking control problem using reinforcement learning (RL) methods.…”

Section: Introductionmentioning

confidence: 99%

Tracking control for networked control systems with DoS attacks via reinforcement learning method

Liu

Dong

Zha

et al. 2023

Preprint

View full text Add to dashboard Cite

This paper is concerned with the tracking control problem for a class of networked systems subject to denial-of-service (DoS) attacks using reinforcement learning methods. Taking the effects of DoS attacks into consideration, a novel value function is proposed, which considers the cost of the control input, external disturbance and tracking error. Then, using the structure of the value function, the tracking Bellman equation and Hamilton function are defined. By employing the Bellman optimality theory, the optimal control strategy and the game algebraic Riccati equation (GARE) are solved with the Hamilton function. Next, the desired tracking performance is guaranteed as the solution of the GARE is found. Furthermore, an attacks-based Q-learning algorithm is projected to find the solution to the optimal tracking problem without the system dynamics and the convergence of the Q-learning algorithm is given. Finally, the F-404 aircraft engine system is given to verify the effectiveness of the proposed control strategy.

show abstract

Section: Introductionmentioning

confidence: 99%

Tracking control for networked control systems with DoS attacks via reinforcement learning method

Liu

Dong

Zha

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Apart from the advanced control algorithm, dynamic model of system is also the basis of precise control performance, but obtaining the accurate model is extremely difficult in mathematics. Artificial neural network (ANN) has long been used as a powerful tool in control of highly nonlinear system [33,34], since the system dynamics can be approximated by it [35]. Thus, for the underwater manipulator subjected to highly unknown system dynamics and coupling effects, the ANN technique is applicable to approximate the uncertainties in the system during the control loop in real-time [36].…”

Section: Introductionmentioning

confidence: 99%

Adaptive super‐twisting control of underwater intervention system considering dynamic couplings and uncertainties

Xiong

Xia

Zhang

et al. 2022

IET Control Theory & Appl

View full text Add to dashboard Cite

Underwater vehicle‐manipulator system (UVMS) is an emerging advanced equipment in underwater intervention scenarios. Several challenges limiting the performance of UVMS's manipulator are treated as burning issues, such as dynamic coupling effects, system uncertainties, disturbances etc. In this article, a novel lightly computational adaptive neural network based backstepping super‐twisting sliding mode control framework is proposed for underwater manipulator trajectory control in UVMS intervention scenario. First, a backstepping super‐twisting sliding mode control (BSSTSMC) scheme is derived, in which the chattering phenomenon of traditional sliding mode control is reduced effectively. Compared with other chattering‐free methods, the higher order control theory maintains the system robustness as well as control accuracy better. Second, a modified neural network (NN) is introduced to predict the unknown system dynamics with the consideration of coupling effects in real‐time. Then, a dimension compression strategy (DCS) for the NN's input layer is proposed to reduce the computational burden and improve the predict performance. Next, based on the Lyapunov method, the system stability is proved, in which the related error variables are guaranteed to be uniformly bounded. Finally, numerical simulations demonstrate the effectiveness and advantages of the proposed DCS‐NN‐BSSTSMC framework through the comparison with other controllers.

show abstract

Adversarial deep reinforcement learning based robust depth tracking control for underactuated autonomous underwater vehicle

Wang,

Xiang,

Duan

et al. 2024

Engineering Applications of Artificial Intelligence

View full text Add to dashboard Cite

Learning‐based robust control methodologies under information constraints

Cited by 3 publications

References 34 publications

Tracking control for networked control systems with DoS attacks via reinforcement learning method

Tracking control for networked control systems with DoS attacks via reinforcement learning method

Adaptive super‐twisting control of underwater intervention system considering dynamic couplings and uncertainties

Adversarial deep reinforcement learning based robust depth tracking control for underactuated autonomous underwater vehicle

Contact Info

Product

Resources

About