In this paper, a new data-based Q-learning algorithm is proposed to address the optimal control issue for a class of discrete-time switched affine systems (SASs). The algorithm shifts the emphasis onto learning the optimal switching law directly from system input-output data, employing a neural-network-approximated Q-function as the key learning element. Firstly, the optimal control issue is transformed into solving the corresponding Bellman’s optimality equation based on the Q-function. Then, a new Q-learning algorithm is developed to find the optimal solution of system switching based entirely on the system input-output data, and a fully connected neural network is borrowed as the Q-function approximator. Moreover, considering the affine properties of SASs, the sequence of Q-functions generated remains bounded in proximity to the precise optimal solution. Finally, both the advantage and effectiveness of the proposed Q-learning based optimal control approach are verified by three examples, including a case study of DC-DC buck-boost converter.

This work addresses the adaptive prescribed-time performance platoon control (PTP-PC) problem for heterogeneous connected autonomous vehicles, which faces two key challenges: invading vehicle tracking and unknown overall disturbance. An information protection spacing policy is proposed to tackle the issue of tracking platoon by the invading vehicle. Besides, the variable gain nonlinear extended state observers (NESOs) and variable gain hyperbolic tangent tracking differentiators (HTTDs) are designed to effectively address the overall disturbance and complexity explosion issues, respectively. To obtain the specified transient and steady-state performance for the entire platoon, a prescribed-time performance function is established. On the basis of the above presented techniques, a distributed adaptive PTP-PC scheme is developed to ensure the vehicular bistability and superior prescribed performance. For practical considerations, both numerical simulations and co-simulation experiments based on PRESCAN/SIMULINK have been conducted to simulate a traffic scenario involving the transportation of college entrance examination papers, and the results have verified the feasibility of the developed scheme.

In this paper, the novel prescribed-time dynamic event-triggered control method of an unknown multiplayer nonzero-sum game (MP-NZSG) is designed by using adaptive dynamic programming (ADP). Firstly, a neural network-based identifier is constructed to estimate the unknown system dynamics. Subsequently, a novel ADP-based dynamic event-triggered control approach is advanced to ensure optimality and prescribed-time stability. A critic neural network (NN) is established for each player to approximate the Nash equilibrium solution of the dynamic event-triggered Hamilton-Jacobi-Isaacs (HJI) equation. This network employs a novel weight updating law, based on the experience replay technique, to alleviate the persistence of excitation condition. Furthermore, using the Lyapunov method, the uniform limit boundedness analysis of the neural network approximation error and multiplayer system is validated. Additionally, minimum inter-event time (MIET) is conclusively established to mitigate the notorious Zeno behaviour. Ultimately, the efficacy of the proposed method is rigorously substantiated through comprehensive simulation results. Note to Practitioners—Our research addresses the challenges of multi-component coordinated control, particularly in spacecraft attitude control. To handle these complexities, we propose an innovative adaptive dynamic event-triggered control approach. By integrating adaptive dynamic programming and neural networks, we effectively model and manage unknown system dynamics, enhancing the controller’s adaptability and robustness. Dynamic event-triggered policies are introduced to optimize system performance and reduce computational costs. The ADP-based prescribed time optimal control scheme prioritizes steady-state performance of nonlinear nonaffine systems, ensuring precise task completion within specified timeframes. Additionally, experience replay technology further fortifies the controller’s learning and adaptability to dynamic environments.

The twin delayed deep deterministic policy gradient (TD3) algorithm and genetic (G) algorithm can take significant time to converge. Hence, it would be interesting to propose an alternative algorithm for fast gains learning in a high-gain controller, being reflected as fast trajectory tracking. In a differential evolution (DE) algorithm, the population is installed, and the mutation, crossover, and selection operations are repeated until the convergence is located. In this way, compared with the TD3 and G algorithms, a DE algorithm can converge faster. In this article, the fast gains learning in a DE high-gain controller (DEHGC) is proposed. The DEHGC contains a high-gain controller for trajectory tracking and a DE algorithm for fast gains learning. The error stability of the high-gain controller is assured. The pseudocode of the DEHGC is detailed. The DE, TD3, and G algorithms are compared for fast gains learning in the high-gain controller.

In this article, we concentrate on the stability and stabilization issues for the discrete-time Takagi-Sugeno (T-S) fuzzy semi-Markov jump systems (FS-MJSs) with incomplete transition probabilities (TPs). Since the statistical properties of mode transitions are frequently hard to get in practice, the system is built on the assumption that the information of TPs is partly accessed. Then, by making use of the incompletely discrete-time semi-Markov kernel, we offer the sufficient requisites for FS-MJSs mean square stability (MSS). We emphasize the impact of TPs on FS-MJSs by utilizing known TPs and introducing them into sufficient conditions for MSS. Numerically testable stability criteria for FS-MJSs in the sense of MSS are provided. The existence criteria are suggested to design the controller to guarantee the MSS of the closed-loop FS-MJSs. At the same time, a homogeneous polynomial technique is developed to reduce conservatism. The numerical examples demonstrate the effectiveness of our method.

Based on the general structure of mechanical systems described by their state-space representation, the Takagi–Sugeno fuzzy modeling, and the controllability property of fuzzy systems, an algebraic and practical approach to computing the fuzzy gain capable of ensuring the stability property of the Takagi–Sugeno fuzzy model is proposed in this article. The main idea consists of finding a continuous fuzzy gain such that any linear behavior, defined by the adequate selection of eigenvalues, is induced in the closed-loop fuzzy system. Therefore, by continuity, if the fuzzy model is an approximation sufficiently close to the mechanical system, then such a nonlinear system is also stabilized by the fuzzy controller. A notable advantage of the proposed method, when compared with similar approaches, is the simplicity of the resulting gain. The validity of the approach is illustrated through the numerical simulation of a sufficiently complex nonlinear system.

The Levenberg-Marquardt and Newton are two algorithms that use the Hessian for the artificial neural network learning. In this article, we propose a modified Levenberg-Marquardt algorithm for the artificial neural network learning containing the training and testing stages. The modified Levenberg-Marquardt algorithm is based on the Levenberg-Marquardt and Newton algorithms but with the following two differences to assure the error stability and weights boundedness: 1) there is a singularity point in the learning rates of the Levenberg-Marquardt and Newton algorithms, while there is not a singularity point in the learning rate of the modified Levenberg-Marquardt algorithm and 2) the Levenberg-Marquardt and Newton algorithms have three different learning rates, while the modified Levenberg-Marquardt algorithm only has one learning rate. The error stability and weights boundedness of the modified Levenberg-Marquardt algorithm are assured based on the Lyapunov technique. We compare the artificial neural network learning with the modified Levenberg-Marquardt, Levenberg-Marquardt, Newton, and stable gradient algorithms for the learning of the electric and brain signals data set.

As you open this January issue of the IEEE Transactions on Neural Networks and Learning Systems (TNNLS), I hope everyone enjoyed a great holiday season and is excited for the new year of 2019. I am very delighted and honored to report several key metrics of IEEE TNNLS to the community.

An approach to design stabilizers and observers for a class of multiple-input multiple-output (MIMO) Takagi-Sugeno (T-S) fuzzy models is developed on the basis of local gains and the searching for a set of interpolation functions capable of properly combining the aforementioned local gains. As expected, the existence of such interpolation functions depends on the controllability and observability properties of the overall multivariable T-S fuzzy model. For that reason, practical controllability and observability tests are also proposed for MIMO T-S fuzzy systems. Some numerical simulations are used in order to validate the efficacy of the method. Besides, the results are compared with an approach based on linear matrix inequalities, namely parallel distributed compensation.

Salt and Pepper noise removal is an important image preprocessing task, it has two simultaneous demands: the suppression of impulses and the preservation of edges. To address this problem in gray scale images, we propose an efficient method which consists of introducing a Redescending M-Estimator w ithin of the Median-Estimator scheme. The Redescending M-Estimator controls the magnitude of the Salt or Pepper impulses and deletes them when it is necessary; the remaining pixels in the neighborhood are processed by the Median-Estimator in order to obtain an estimation of a noise free pixel. The proposed scheme is applied on the entire image using sliding windows of size 5 × 5; the local information obtained by this window is used to calculate the thresholds and the parameters that characterize the influence functions tested in the Redescending M-Estimator. To improve the suppression ability of our proposal a pulse detector is used, it identifies when is necessary to submit each pixel to the denoising process. The effectiveness of our proposal is verified by quantitative and qualitative results.

This paper is concerned with the problem of asynchronous H∞filtering for a class of discrete-time Takagi-Sugeno fuzzy affine systems against time-varying signal transmission delays and measurement quantization. The asynchrony refers to the situation that the plant state and the filter state belong to different local state space regions, and the quantization density can be adjusted to satisfy different performance requirements at different time instants. By transforming the filtering error system into an input-output form consisting of two interconnected subsystems, sufficient conditions on the existence of the desired asynchronous filter are established via the scaled small gain theorem to ensure that the closed-loop system is asymptotically stable with a prescribed H∞performance index with the aid of a novel piecewise Lyapunov-Krasovskii functional and the S-procedure approach. Finally, a practical example of cart-pendulum with a modified model is provided to illustrate the effectiveness of the obtained theoretical results.

Neural networks (NNs) have numerous applications to online processes, but the problem of stability is rarely discussed. This is an extremely important issue because, if the stability of a solution is not guaranteed, the equipment that is being used can be damaged, which can also cause serious accidents. It is true that in some research papers this problem has been considered, but this concerns continuous-time NN only. At the same time, there are many systems that are better described in the discrete time domain such as population of animals, the annual expenses in an industry, the interest earned by a bank, or the prediction of the distribution of loads stored every hour in a warehouse. Therefore, it is of paramount importance to consider the stability of the discrete-time NN. This paper makes several important contributions. 1) A theorem is stated and proven which guarantees uniform stability of a general discrete-time system. 2) It is proven that the backpropagation (BP) algorithm with a new time-varying rate is uniformly stable for online identification and the identification error converges to a small zone bounded by the uncertainty. 3) It is proven that the weights' error is bounded by the initial weights' error, i.e., overfitting is eliminated in the proposed algorithm. 4) The BP algorithm is applied to predict the distribution of loads that a transelevator receives from a trailer and places in the deposits in a warehouse every hour, so that the deposits in the warehouse are reserved in advance using the prediction results. 5) The BP algorithm is compared with the recursive least square (RLS) algorithm and with the Takagi-Sugeno type fuzzy inference system in the problem of predicting the distribution of loads in a warehouse, giving that the first and the second are stable and the third is unstable. 6) The BP algorithm is compared with the RLS algorithm and with the Kalman filter algorithm in a synthetic example.

In this paper, an online self-organizing fuzzy modified least-square (SOFMLS) network is proposed. The algorithm has the ability to reorganize the model and adapt itself to a changing environment where both the structure and learning parameters are performed simultaneously. The network generates a new rule if the smallest distance between the new data and all the existing rules (the winner rule) is more than a prespecified radius. The major contributions of this paper are as follows: 1) A new network is proposed, in which unidimensional membership functions are used, and only two parameters for each rule are employed, thus reducing the number of parameters. The network avoids the singularity produced by the widths in the antecedent part for online learning; 2) a new pruning algorithm based on the density is proposed, where the density is the number of times each rule is used in the algorithm. The rule that has the smallest density (the looser rule) in a selected number of iterations is pruned if the value of its density is smaller than a prespecified threshold; and 3) the stability of the proposed algorithm is proven, and the bound for the average of the identification error is found. The condition that led the algorithm to avoid the local minimum is found, and it is proven that the parameter error is bounded by the initial parameter error. Three simulations give the effectiveness of the suggested algorithm.

Bounding ellipsoid (BE) algorithms offer an attractive alternative to traditional training algorithms for neural networks, for example, backpropagation and least squares methods. The benefits include high computational efficiency and fast convergence speed. In this paper, we propose an ellipsoid propagation algorithm to train the weights of recurrent neural networks for nonlinear systems identification. Both hidden layers and output layers can be updated. The stability of the BE algorithm is proven.

Compared to normal learning algorithms, for example backpropagation, Kalman filter-based algorithm has some better properties, such as faster convergence. In this paper, Kalman filter is modified with a risk-sensitive cost criterion, we call it as risk-sensitive Kalman filter. This new algorithm is applied to train recurrent neural networks for nonlinear system identification. Input-to-state stability is used to prove that the risk-sensitive Kalman filter training is stable. The contributions of this paper are: 1) the risk-sensitive Kalman filter is used for the state-space recurrent neural networks training, 2) the stability of the risk-sensitive Kalman filter is proved.