Search

This paper is concerned with the event-triggered anti-disturbance control of Markov jump systems with general transition probabilities. The associated multiple disturbances cover matched and unmatched cases. Two dynamic triggering mechanisms are constructed by utilizing the tanh -function to adjust thresholds varying with input error. An adaptive disturbance observer is presented in terms of a row-by-row configuration to estimate unknown matched disturbance. According the mechanisms, corresponding composite state-feedback controllers are proposed by integrating threshold bound and adaptive estimation. Resorting to the Lyapunov stability theory and the stochastic analysis technique, the resulted closed-loop system is stochastically bounded with the required H ∞ performance. A structured separation method is utilized to solve the controller gain in terms of linear matrix inequalities. Finally, the validity of proposed schemes is verified by a numerical simulation comparison. [ABSTRACT FROM AUTHOR]

This paper is dedicated to a proportional–integral intermediate estimator based fault-tolerant control method for Markov jump systems with event-triggered inputs. A proportional–integral type intermediate estimator is established to reconstruct sensor and actuator faults, as well as system states. A dynamic event-triggered mechanism is designed to reduce unnecessary transmissions between the controller and the actuator. A threshold-dependent fault-tolerant controller is established to actively compensate for actuator faults. Finsler's lemma is adopted to decouple the control gain, the observer gain and the Lyapunov variable. At last, the proposed estimation method is verified by a comparison example. • A PI–type intermediate observer is designed to obtain better estimation performance. • A threshold-dependent triggering scheme is designed to reduce the network burden. • A structured separation method is adopted to address complex couplings between gains. [ABSTRACT FROM AUTHOR]

• The uncertainty is a mismatched type without known upper bounds. • Three triggering mechanisms are proposed to reduce the redundant control signal transmission. • The corresponding triggering controllers containing triggering thresholds bounds are constructed to compensate the uncertainties and ensure the required H ∞ performance. This paper investigates the H ∞ control of uncertain linear systems under the event-triggered control framework. The uncertainty is a mismatched type without known upper bounds. To reduce the redundant transmission of control signals, three triggering strategies are developed, namely, the absolute threshold triggering manner is built firstly to deal with the small control input variation. By revisiting the first one, the large control input is monitored by the second triggering scheme. Lastly, a trade-off between the above two triggering measures is made to cover all possible inputs without sacrificing system performance largely. Based on the proposed triggering strategies, different adaptive controllers containing the triggering threshold bound are constructed to eliminate the unknown uncertainty and ensure uniformly bounded with the required H ∞ performance level, which is verified by simulation studies. [ABSTRACT FROM AUTHOR]

In recent years, the deep reinforcement learning (DRL) algorithms have been developed rapidly and have achieved excellent performance in many challenging tasks. However, due to the complexity of network structure and a large amount of network parameters, the training of deep network is time-consuming, and consequently, the learning efficiency of DRL is limited. In this paper, aiming to speed up the learning process of DRL agent, we propose a novel approximate policy-based accelerated (APA) algorithm from the viewpoint of the error analysis of approximate policy iteration reinforcement learning algorithms. The proposed APA is proven to be convergent even with a more aggressive learning rate, making the DRL agent have a faster learning speed. Furthermore, to combine the accelerated algorithm with deep Q-network (DQN), Double DQN and deep deterministic policy gradient (DDPG), we proposed three novel DRL algorithms: APA-DQN, APA-Double DQN, and APA-DDPG, which demonstrates the adaptability of the accelerated algorithm with DRL algorithms. We have tested the proposed algorithms on both discrete-action and continuous-action tasks. Their superior performance demonstrates their great potential in the practical applications. [ABSTRACT FROM AUTHOR]