Radius Basis Function Neural Network Research Articles

Adaptive Asymmetric Time-Varying Integral Barrier Lyapunov Control-Based Trajectory Tracking of Autonomous Vehicles

This paper investigates the lane following and changing maneuvers of autonomous vehicles in the presence of unknown disturbances, taking into account the dynamic system states and input constraints. The integrated longitudinal-lateral and yaw rate dynamics of the vehicle are simultaneously considered to improve the tracking accuracy and system stability when navigating under critical conditions. Then, an adaptive asymmetric time-varying integral barrier Lyapunov control and dynamic surface control scheme are developed to design the active front steering controller, longitudinal controller, and direct yaw moment control controller, which is capable of constraining the system states and control signals within the predefined boundary. In addition, the radius basis function neural network (RBFNN) is employed to estimate the lumped disturbances caused by the parametric uncertainties, external disturbances, and unmodeled dynamics, and the command filter system is used to avoid the explosion of terms phenomenon. Due to the fast and accurate torque response characteristics of the in-wheel motors, the optimization-based method is then implemented to effectively allocate the driving/braking torque to each in-wheel motor so as to improve vehicle performance. The stability of the closed-loop system is comprehensively demonstrated by means of the Lyapunov theory. Finally, the quantitative and qualitative comparisons in different diving scenarios using the Carim-Simulink joint environment are carried out to illustrate the effectiveness and validation of the proposed method.

Model‐free hierarchical control with fractional‐order sliding surface for multisection web machines

AbstractDue to the nonlinearities and dynamical properties of multisection web machines, it is essential to accomplish an appropriate mathematical model as well as design a compatible control scheme. A generalized fully driven web winding model is constructed taking unpredictable changes of moment of inertia and radius of rolls into consideration. In this article, a fractional‐order sliding surface algorithm‐based hierarchical control frame is applied to multisection web machines to deal with matched and mismatched uncertainties and disturbances. In addition, the Radius Basis Function neural network is deployed to estimate necessary system dynamics and the proposed control scheme taking advantage of first‐order low‐pass filters to produce virtual control signals without complicated calculation and attenuate the explosion of terms phenomenon. Finally, the semi‐global stability of the entire system is determined using Lyapunov's stability theory. The simulation results are presented to clarify the superior performance of robust control algorithm and approximation law in the presence of nonideal elements.

DSC-based RBF neural network control for nonlinear time-delay systems with time-varying full state constraints

The presented control scheme in this paper aims at stabilizing uncertain time-delayed systems requiring all states to change within the preset time-varying constraints. The controller design framework is based on the backstepping method, drastically simplified by the dynamic surface control technique. Meanwhile, the radius basis function neural networks are utilized to deal with the unknown items. To prevent all state variables from violating time-varying predefined regions, we employ the time-varying barrier Lyapunov functions during the backstepping procedure. Moreover, appropriate Lyapunov–Krasovskii functionals are used to cancel the influence of the time-delay terms on the system’s stability. Under the presented control laws and Lyapunov analysis, it is proven that constraints on all state variables are not breached, good tracking performance of desired output is achieved, and all signals in the closed-loop systems are bounded. The effectiveness of our control scheme is confirmed by a simulation example.

Cost-optimal energy management strategy for plug-in hybrid electric vehicles with variable horizon speed prediction and adaptive state-of-charge reference

In this paper, an energy management strategy (EMS) based on model predictive control (MPC) is proposed to minimize fuel cost, electricity usage and battery ageing. To fulfil the MPC framework, a novel speed predictor with a variable horizon based on a K-means algorithm and a radius basis function neural network, which contains various predictive submodels, is designed to cope with different input drive states. In addition, a Q-learning algorithm is applied to construct an adaptive multimode state-of-charge (SOC) reference generator, which takes advantage of velocity forecasts for each prediction horizon. The algorithm fully considers the model nonlinearities and physical constraints and requires less computational effort. Based on the SOC reference and predictive velocity, the MPC problem is formulated to coordinate fuel consumption and battery degradation. Moreover, considering the influence of real-time traffic information, a traffic model that simulates actual road conditions is constructed in VISSIM to evaluate the performance of the proposed EMS. The simulation results show that the proposed speed predictor can effectively improve the predictive accuracy, and the multimode control laws based on drive condition classification present superior adaptability in SOC reference generation compared to single-mode law. With the aforementioned two improvements, the proposed EMS achieves desirable performance in fuel economy and battery lifetime extension.

Robust adaptive anti-windup wheel slip tracking control for intelligent vehicle with fast terminal sliding mode observer

Aiming at the requirement of the intelligent vehicle for the fast and stable tracking control of the wheel slip, a novel robust adaptive anti-windup wheel slip tracking control method with fast terminal sliding mode observer is proposed. First, a fast terminal sliding mode observer based on equivalent control on the sliding surface is proposed to estimate the states of the wheel slip dynamic system to lay the foundation for the full state feedback control law design. Second, a robust adaptive anti-windup wheel slip tracking control law with lumped uncertainty observer and additional anti-windup dynamics is derived based on Lyapunov-based method. The lumped uncertainty observer utilizes the nonlinear mapping ability of the radius basis function neural network to estimate and compensate the lumped uncertainty of the system, and the unknown optimal weight vector of the radius basis function neural network is updated by adaptive law. The additional anti-windup dynamics is used to suppress the effect of the input saturation on the stability of the system. Finally, the performance of the proposed method is verified through simulations of various maneuvers on vehicle dynamics simulation software.

An error similarity-based error-compensation method for measurement in the nonuniform temperature field

Measurement error compensation is a key issue for the application of laser trackers in an unsteady environment, especially in the nonuniform temperature field. Aimed at reducing induced measurement uncertainty, a measurement error compensation method with error similarity analysis is proposed in this paper. The spatial similarity of measurement errors is qualitatively analyzed based on the mathematical model, and quantitative analysis is presented by means of simulation and semivariogram. We then model the error similarity, and an error-estimation model for target points is established with the radius basis function neural network and regularized extreme learning machine. The experiments are conducted to verify error similarity and the applicability and accuracy of the estimation model in practice. The results show that error similarity does exist in practical measurements, and the proposed method can reduce the absolute measurement error to less than 0.03 mm (). The method is applicable for measurements in different environments since the compensated point sets are almost coincident.

A radial‐basis‐function neural network‐based energy reserving strategy for energy‐saving elevator

Efficiency improvement and energy consumption decrease are becoming a pivotal issue in vehicle applications. Predictive control based on a radius‐basis‐function neural network (RBFNN) is employed to develop a new energy‐reserving strategy (ERS) that dynamically regulates the energy stored in the supercapacitor pack (SCP), which is equipped in an energy‐saving elevator system. Specifically, at the beginning of every traveling period, the balance voltage (BV) of the SCP, which represents the instantaneous state of the energy stored, is predicted and managed based on the traveling distance and the load ratio of the elevator. In this way, the capacitance of the SCP can be fully used to provide or reserve as much energy as possible. The above energy optimization problem is modeled and transformed into a group of constrained optimization equations based on the energy analysis of the elevator. With the proposed control scheme, not only does the peak power, which comes out when the elevator moves with heavy load eliminate, but also part of the nominal power can be provided by the SCP. Thus, the power provided by the AC‐grid pulse‐width modulation (PWM) rectifier is decreased, and the energy discharged by the SCP, which has reserved the braking energy at the previous journey, is optimized. The paper includes a detailed explanation of power and energy flow of the elevator and the implementation of the proposed RBFNN‐based predictive control scheme. Finally, several simulation and experimental results are presented to demonstrate the effectiveness of the presented control scheme. © 2018 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

Spot-color matching system of flexo printing based on radius basis function neural network

Spot-color matching system of flexo printing based on radius basis function neural network

Simulation and prediction of the thuringiensin abiotic degradation processes in aqueous solution by a radius basis function neural network model

The thuringiensin abiotic degradation processes in aqueous solution under different conditions, with a pH range of 5.0–9.0 and a temperature range of 10–40°C, were systematically investigated by an exponential decay model and a radius basis function (RBF) neural network model, respectively. The half-lives of thuringiensin calculated by the exponential decay model ranged from 2.72d to 16.19d under the different conditions mentioned above. Furthermore, an RBF model with accuracy of 0.1 and SPREAD value 5 was employed to model the degradation processes. The results showed that the model could simulate and predict the degradation processes well. Both the half-lives and the prediction data showed that thuringiensin was an easily degradable antibiotic, which could be an important factor in the evaluation of its safety.

Ultra-spectra communication system (USCS): a promising way to energy-efficient green communication networks

In this article, we investigate a promising transmission approach for green communication networks. By combining the appealing single-cycle modulation technique with traditional pulse modulation found on the underlay communication philosophy, the presented ultra-spectra communication system can improve the energy efficiency significantly. In transceiver, the emission signal contains two parts, the information-bearing baseband waveform with an extremely short duration and the information-synchronized carrier component. Alternatively, without carrier signaling the intermittent emission can be adopted to further reduce its operational power. By optimally allocating the power in different bands, a water-filling emission spectrum inspired by a novel conception of radius basis function neural-network is proposed for maximizing transmission energy efficiency in the presence of external interference. In receiver, the carrier-synchronized signal greatly alleviates the power-consumed timing acquisition, and provides extremely precise synchronization for information detection in intensive multipath environments. In addition, from a pattern classification perspective, a low-complexity noncoherent detection scheme is suggested to enhance receiving performance and minimize energy consumption. Performance analysis demonstrated that the whole transmission scheme can indeed be realized within an ultra-low-power framework, which shows great advantages to traditional strategies.

Adaptive Neural Control for a Class of Outputs Time‐Delay Nonlinear Systems

This paper considers an adaptive neural control for a class of outputs time‐delay nonlinear systems with perturbed or no. Based on RBF neural networks, the radius basis function (RBF) neural networks is employed to estimate the unknown continuous functions. The proposed control guarantees that all closed‐loop signals remain bounded. The simulation results demonstrate the effectiveness of the proposed control scheme.

Estimation of biochemical variables using quantum- behaved particle swarm optimization (QPSO)-trained radius basis function neural network: A case study of fermentation process of L-glutamic acid

Due to the difficulties in the measurement of biochemical variables in fermentation process, softsensing model based on radius basis function neural network had been established for estimating the variables. To generate a more efficient neural network estimator, we employed the previously proposed quantum-behaved particle swarm optimization (QPSO) algorithm for neural network training. The experiment results of L-glutamic acid fermentation process showed that our established estimator could predict variables such as the concentrations of glucose, biomass and glutamic acid with higher accuracy than the estimator trained by the most widely used orthogonal least squares (OLS). According to its global convergence, QPSO generated a group of more proper network parameters than the most popular OLS. Thus, QPSO-RBF estimator was more favorable to the control and fault diagnosis of the fermentation process, and consequently, it increased the yield of fermentation. Key words: Soft-sensing model, quantum-behaved particle swarm optimization algorithm, neural network.

Research on Active Suspension System Based on BP and RBF Network Algorithm

A six degrees of freedom half body vehicle suspension system is presented in the paper .The Back Propagation neural network algorithm and the Radial-Basis Function network algorithm is adopted to control the suspension system. With the aid of software Matlab/Simulink , the simulation model is obtained. A great deal of simulation work is done. Simulation results demonstrate that both the designed radius basis function neural network and the back propagation neural network work well for the proposed vehicle suspension model in the paper .

Attribute dependency theory and its application on neural network

Neural network optimization methods are generally confined to learning algorithms and input attributes.Due to the high-dimensional mapping that neural network fits contains complex intrinsic attribute dependencies,the traditional optimization methods have not conducted the analytical study on it.The article put forward the attribute dependency theory based on functional dependency theory,elaborated the definition of the attribute dependency theory,and proved related theorem.Combining the Radius Basis Function(RBF)neural network,a new neural network optimization method based on attribute dependency theory(ADO-RBF)was proposed.

Interference Mitigation between Ultra-Wideband Sensor Network and Other Legal Systems

Ultra-wideband impulse radio (UWB-IR) sensor network has intensive military and commercial applications. However, the interference between UWB and other existed networks should be casually investigated. In this paper, we consider interference mitigation in UWB sensors in the context of cognitive radio (CR). Firstly, we suggest a general state transition model to characterize the working states evolution of legal networks, also referred to as primary users (PU). Spectrum sensing, used to identify the state of PU, is formulated as detection of a corresponding state sequence. Maximum posterior probability (MAP) criterion is adopted to perform spectrum sensing. By exploring potential gain of state transitions, detection probability for nearby networks is improved significantly. Subsequently, based on the radius basis function neural network (RBF), we present a novel spectrum sculptor to design UWB waveforms. Attributed to the excellent reconfiguration of RBF, our scheme can produce UWB waveforms tracing available spectrums. The designed waveforms can entirely utilize multiple unoccupied bands to maintain uninterrupted communications. Also, sufficient spectral attenuation can be generated in specific bands to mitigate mutual interference between UWB sensors and other networks. Besides, orthogonal waveforms can be easily derived, which either improves transmission performance or provides a flexible accessing strategy for UWB sensors.

Adaptive RBF neural-networks control for a class of time-delay nonlinear systems

A control scheme combined with backstepping, radius basis function (RBF) neural networks and adaptive control is proposed for the stabilization of nonlinear system with input and state delay. By using state transformation, the original system is converted to the system without input delay. The RBF neural network is employed to estimate the unknown continuous function. The controller is designed for the converted system so that the closed-loop system is bounded. According to the relation between the original system and the converted one, the state of the original system is proved to be bounded. The control scheme ensures that the closed-loop system is semi-globally uniformly ultimately bounded.

DECOUPLING NEURAL SLIDING MODE CONTROL FOR MULTI-LINK ROBOTS

A decoupling neural sliding mode controller is given for trajectory tracking control of multi-link robots with external disturbances and uncertain system parameter errors. Different from general combining the sliding mode control (SMC) and neural network control (NNC), this approach decouples a robot with n links into n subsystems and the state response of each subsystem can be governed by a corresponding sliding mode manifold. The whole system is controlled by a whole sliding mode manifold selected to be a global fast terminal sliding mode manifold, which guarantees that the controlled system can reach the sliding mode manifold and equilibrium point in finite time. A radius basis function (RBF) neural network is applied to learn the limit of system parameter errors and external disturbances for every subsystem, and enforce sliding mode motion by minimizing the cost function that is with respect to the distance from the sliding mode manifold. The switching gain of sliding mode control can be automatically adjusted according to system parameter errors and external disturbances by RBF neural network's learning. The controller's chattering is reduced without sacrificing robustness, even eliminated after RBF neural network's learning for a short time. Moreover, the stability of the controlled system is proven and the convergence of tracking error is also proven by Lyapnov function. Simulation results verify the performance of the control scheme.

Identification between Stephania tetrandra S. Moore and Stephania cepharantha Hayata by CWT–FTIR–RBFNN

Horizontal attenuation total reflection-Fourier transform infrared spectroscopy (HATR-FTIR) is used to measure the FTIR ofStephania tetrandra S. Moore andStephania cepharantha Hayata. Because they belong to the same family and the same genus Chinese traditional medicinal materials, their chemical components are very similar. In order to extrude the difference between Stephania tetrandra S. Moore and Stephania cepharantha Hayata, continuous wavelet transform (CWT) is used to decompose the FTIR of Stephania tetrandra S. Moore and Stephania cepharantha Hayata. Three main scales are selected as the feature extracting space in the CWT domain. According the distribution of FTIR of Stephania tetrandra S. Moore and Stephania cepharantha Hayata, three feature regions are determined at every spectra band at selected three scales in the CWT domain. Thus nine feature parameters form the feature vector. The feature vector is input to the radius basis function neural network (RBFNN) to train so as to accurately classify the Stephania tetrandra S. Moore and Stephania cepharantha Hayata. 128 couples of FTIR are used to train and test the proposed method, where 78 couples of data are used as training samples and 50 couples of data are used as testing samples. Experimental results show that the accurate recognition rate between Stephania tetrandra S. Moore and Stephania cepharantha Hayata is respectively 99.8 and 99.9% by using the proposed method.

Identification betweenStephania tetrandraS. Moore andStephania cepharanthaHayata by CWT–FTIR–RBFNN

Horizontal attenuation total reflection–Fourier transform infrared spectroscopy (HATR–FTIR) is used to measure the FTIR ofStephania tetrandraS. Moore andStephania cepharanthaHayata. Because they belong to the same family and the same genus Chinese traditional medicinal materials, their chemical components are very similar. In order to extrude the difference betweenStephania tetrandraS. Moore andStephania cepharanthaHayata, continuous wavelet transform (CWT) is used to decompose the FTIR ofStephania tetrandraS. Moore andStephania cepharanthaHayata. Three main scales are selected as the feature extracting space in the CWT domain. According the distribution of FTIR ofStephania tetrandraS. Moore andStephania cepharanthaHayata, three feature regions are determined at every spectra band at selected three scales in the CWT domain. Thus nine feature parameters form the feature vector. The feature vector is input to the radius basis function neural network (RBFNN) to train so as to accurately classify theStephania tetrandraS. Moore andStephania cepharanthaHayata. 128 couples of FTIR are used to train and test the proposed method, where 78 couples of data are used as training samples and 50 couples of data are used as testing samples. Experimental results show that the accurate recognition rate betweenStephania tetrandraS. Moore andStephania cepharanthaHayata is respectively 99.8 and 99.9% by using the proposed method.

An alternative approach using artificial neural networks for power transformer protection

AbstractThis work presents an alternative approach using the differential logic associated to artificial neural networks (ANNs) in order to distinguish between inrush currents and internal faults for the protection of power transformers. The radius basis function (RBF) neural network is proposed as an alternative approach in order to distinguish the situations described, using a smaller amount of data for training purposes, in some cases, if compared with networks such as the multi‐layer perceptron (MLP). The ANN results are then compared to those obtained by the traditional differential protection algorithm. An ANN approach for correction of saturated current signals is also presented. Copyright © 2005 John Wiley & Sons, Ltd.