Search

This paper investigates the global asymptotic stability problem of recurrent neural networks with multiple time-varying delays. First, to make full use of the relationship between all delayed states x ( t - τ i ) ( i = 1 , … , N ) and current state x ( t ) , an augmented delays-dependent region partitioning (ADRP) approach is proposed. Then combining Wirtinger-based integral inequality and the reciprocally convex approach, two delay-dependent stability criteria with less conservatism are developed. Finally, two numerical simulations are shown to verify the effectiveness and the feasibility.

Transfer neural networks have been successfully applied in many domain adaptation tasks. The initiative of most of the current transfer networks, essentially, is optimizing a single distance metric between the source domain and target domain, while few studies integrate multiple metrics for training transfer networks. In this paper, we propose an architecture of transfer neural network equipped with hybrid representations of domain discrepancy, which could incorporate the advantages of different types of metrics as well as compensate their imperfections. In our architecture, the Maximum Mean Discrepancy (MMD) and H -divergence based domain adaptations are combined for simultaneous distribution alignment and domain confusion. Through extensive experiments, we find that the proposed method is able to achieve compelling transfer performance across the datasets with domain discrepancy from small scale to large scale. Especially, the proposed method can be promisingly used to predict the viewpoint of 3D-printed workpiece even trained without labels of real images. The visualization of learned features and adapted distributions by our transfer network highlights that the proposed approach could effectively learn the similar features between two domains and deal with a wide range of transfer tasks.

Adaptive orthogonal frequency-division multiplexing (OFDM) is a promising technology for underwater acoustic sensor networks (UASNs) to facilitate robust and reliable transmission. This paper deals with an adaptive UASN-OFDM multi-parameter allocation problem in a strongly incomplete information scenario. Specifically, an adversarial multi-armed bandit (MAB) formalism is first proposed, whereby no prior knowledge about channel conditions is required and the reward sequences are not restrained by any statistical assumptions. Second, considering the curse of dimensionality caused by exponentially large number of feasible strategies, we tailor orthogonal learning strategy to reinforce learning for initial decision set and achieve filtration by abandoning some inferior levels. Third, under strictly limited prior information, we design a time-based dynamic exploration mechanism to adjust exploration factor adaptively, which improves algorithm learning ability effectively. Thank to aforementioned efforts, a low-complexity, high-efficiency OD-Exp3 algorithm is presented to handle the complex adaptive OFDM problem in UASNs. Lastly, we show the upper regret bound and the convergence of OD-Exp3 algorithm. Comparative results demonstrate that the proposed algorithm is superior to the existing algorithms.

This paper proposed an online image-based visual servoing (IBVS) controller for manipulator systems with dead-zone input. The adaptive neural networks (NNs) are used to approximate the unknown nonlinear dynamics. The Barrier Lyapunov Function (BLF) is constructed to overcome the visibility constraint problem, in which both the constant symmetric barriers and time-varying asymmetric barriers are considered. With the proposed control method, it is proved that all the signals in the closed-loop system are semi-globally uniformly bounded and the image error is remained in a bounded compact set. Finally, simulation examples are given to illustrate the effectiveness of the proposed control method.

In this paper, the distributed bearing-based formation control problem for networked robotic systems with parametric uncertainties is investigated. Firstly, under the consideration that the task-space velocity is measurable, a reference control input is designed to achieve a bearing constrained target formation. For the unmeasurable task-space velocity case, an observer-based reference velocity scheme is proposed and only the local relative task-space position measurement is needed to achieve globally bearing-based formation stabilization. By designing a velocity feedback in proportional-integral reference velocity control scheme, at least two leaders can handle the leader–follower formation tracking problem, in which the followers do not need any global information. Finally, some simulation results are provided to demonstrate the effectiveness of the proposed control laws.

This study considers a joint channel and power allocation for multiple users in underwater acoustic communication networks as a formulated multiplayer MAB game. This study also proposes hierarchical learning algorithms, which do not need any prior environmental information and direct information exchange among users, to improve the learning ability. In upper sub-learning, each user generates a strategy through the traditional UCB1 strategy. In lower sub-learning, the concept of virtual learning information, which can be obtained as the reward of the last actual played strategy, is introduced to enrich the learning information. Users can enhance their learning ability by learning the outdated virtual learning information in lower sub-learning. As a result, the learning time it takes to achieve the NE is effectively decreased, and the cost of the algorithm is reduced. A distributed optimal NE selection mechanism is proposed to avoid falling into an inadequate local extreme value. Simulation results show high convergence speed and achieved utility of the proposed algorithm.

This paper focuses on the problem of decentralized event-triggered control for a class of interconnected time delay stochastic nonlinear systems with unmodeled dynamics. In order to ensure that the outputs of systems satisfie the prescribed performance, a funnel performance variable is introduced. By using neural network approximation theory and backstepping method, the controller and its event triggered mechanism are co-designed. Based on Lyapunov stability theory and changing supply functions idea, it is proved that all the signals of the overall closed-loop system with the designed controller are bounded in probability. Finally, a numerical simulation is presented to illustrate the effectiveness of the proposed method.

In this paper, the robust delay-dependent stability problem is investigated for discrete-time recurrent neural networks (DRNNs) with time-varying delays and nonlinear perturbations. A novel summation inequality is proposed, which takes information on the double summation of system state into consideration and further extends the discrete Wirtinger-based inequality. By utilizing technique of the novel inequality and Lyapunov-Krasovskii functionals, a sufficient condition on robust stability of DRNNs with time-varying delays and nonlinear perturbations is obtained in terms of linear matrix inequality. The numerical example is included to show that the proposed method is effective and provides less conservative results.

The adaptive tracking control problem is considered for a class of nonlinear time-delay systems in the presence of input and tracking error constraint. A reduced-order observer is designed to estimate the unmeasured state variables at first. Then, a constraint variable is utilized to ensure that the tracking error is within the prescribed boundaries. An auxiliary state is introduced to deal with the input saturation constraint. With the time-delay functions unavailable, we employ adaptive RBF neural network systems to approximate unknown functions. It is proved that the resulting closed-loop system is stable in the sense of semiglobal uniformly ultimately boundedness. The simulations are performed and the results demonstrate the effectiveness of the proposed approach.

Extreme Learning Machine (ELM), a competitive machine learning technique for single-hidden-layer feedforward neural networks (SLFNNs), has proven to be efficient and effective algorithm for regression and classification problems. However, traditional ELM involves a large number of hidden nodes for complex real world regression and classification problems which increasing the computation burden. In this paper, a decomposition based fast ELM (DFELM) algorithm is proposed to effectively reduce the computational burden for large number of hidden nodes condition. Compared with ELM algorithm, DFELM algorithm has faster training time with a large number of hidden nodes maintaining the same accuracy performance. Experiment on three regression problems, six classification problems and a complex blast furnace modeling problem are carried out to verify the performance of DFELM algorithm. Moreover, the decomposition method can be extended to other modified ELM algorithms to further reduce the training time.

In this paper, a novel algorithm is developed for identifying Hammerstein model. The static nonlinear function is characterized by function link artificial neural network (FLANN) and the linear dynamic subsystem by an ARMA model. The utilization of FLANN can not only result in a simple and effective representation of static nonlinearity but also simplify the learning algorithm. A two-step procedure is adopted to identify Hammerstein model by using a specially designed input signal, which separates the identification of linear part from that of nonlinear part. Levenberg-Marquart algorithm is used to learn the weights of FLANN. Simulation examples demonstrate the effectiveness of the proposed method.

A new neural network observer-based networked control structure for a class of nonlinear systems is developed and analyzed. The structure is divided into three parts: local linearized subsystem, communication channels and remote predictive controller. A neural-network-based adaptive observer is presented to approximate the state of the time-delay-free nonlinear system. The neural-network (NN) weights are tuned on-line and no exact knowledge of nonlinearities is required. The time delays considered in the forward and backward communication channels are constant and equal. A modified Smith predictor is proposed to compensate the time delays. The controller is designed based on the developed NN observer and the proposed Smith predictor. By using the Lyapunov theory, rigorous stability proofs for the closed-loop system are presented. Finally, simulations are performed and the results show the effectiveness of the proposed control strategy.

This paper studies the problem of output feedback control for interconnected time-delay systems with prescribed performance. Currently, few of the existing results consider the prescribed performance control in the nonlinear interconnected time-delay systems. The subsystems are in the form of triangular structure with unmodeled dynamics. First, we design a reduced-order observer to estimate the unmeasured states online instead of the traditional full-order observer. Then, by proposing a new state transformation with the performance function, we construct a novel output feedback controller with the idea of the backstepping method. It is strictly proved that the resulting closed-loop system is stable in the sense of uniformly ultimately boundedness and both transient and steady-state performances of the outputs are preserved. Finally, a simulation example is given and the results show the effectiveness of the proposed control design method.

The trajectory tracking problem for the teleoperation systems is addressed in this paper. Two neural network-based controllers are designed for the teleoperation system in free motion. First, with the defined synchronization variables containing the velocity error and the position error between master and slaver, a new adaptive controller using the acceleration signal is designed to guarantee the position and velocity tracking performance between the master and the slave manipulators. Second, with the acceleration signal unavailable, a controller with the new synchronization variables is proposed such that the trajectory tracking error between the master and the slave robots asymptotically converges to zero. By choosing proper Lyapunov functions, the asymptotic tracking performance with these two controllers is proved without the knowledge of the upper bound of the neural network approximation error and the external disturbance. Finally, the simulations are performed to show the effectiveness of the proposed methods.

Antilock braking system (ABS) is a highly nonlinear system including variation and uncertainties in the parameters due to changes in vehicle loadings, road condition, etc. It is a difficult task to design an ideal controller for ABS. In this paper, a novel robust controller named fractional order sliding mode controller (FOSMC) is proposed for ABS to regulate the slip to a desired value. The proposed FOSMC combines sliding mode controller (SMC) with fractional order dynamics, in which fractional order proportional-derivative (FOPD) sliding surface is adopted. FOSMC can not only deal with the uncertainties in ABS system but also track the desired slip faster than conventional integer order SMC with proportional or proportional-derivative sliding surface. Experimental results demonstrate the effectiveness of the proposed method.

This paper presents tracking control problem of the unmatched uncertain nonlinearly parameterized systems (NLP-systems) with unknown input nonlinearities. Two kinds of nonlinearities existing in the control input are discussed, which are non-symmetric dead-zone input and continuous nonlinearly input. The smooth controller is proposed in either of these two cases by effectively integrating adaptive backstepping technique and neural networks. Some assumptions, in which the parameters with respect to the input nonlinearities are available in advance in previous works, are removed by adaptive strategy. The researches also take the arbitrary unmatched uncertainties and nonlinear parameterization into account without imposing any condition on the system. It is shown that the closed-loop tracking error converges to a small neighborhood of zero. Finally, numerical examples are initially bench tested to show the effectiveness of the proposed results.