Recurrent High Order Neural Network Research Articles

Neural control of an induction motor with regenerative braking as electric vehicle architecture

This paper presents the synthesis of an induction motor neural controller and a regenerative braking controller for an electric vehicle architecture, based on two energy system, a Main Energy System (MES) and an Auxiliary Energy System (AES). Such controllers are based on system identification, trajectory tracking and state estimation. System identification uses a Recurrent High Order Neural Network (RHONN), trained with an Extended Kalman Filter (EKF). RHONN obtains an accurate motor model which is robust in presence of external disturbances and parameter variations. To force the motor to track a desired torque trajectory and to reject undesired disturbances, an inverse optimal controller based on the identified neural model is proposed. Therefore, the proposed scheme does not need a-priori knowledge of motor parameters. For state estimation a super-twisting observer is implemented to estimate the rotor magnetic fluxes. The regenerative braking controller addresses the issue when the battery is not capable to accept the generated energy due to braking; Therefore, an AES based on a super capacitor and a buck–boost converter is a solution to recover the braking energy and give power to the motor during acceleration. The regenerative braking controller is based on a PI control to regulate voltage and current of the super capacitor. Simulation and experimental results illustrate the performance of the proposed controllers which are implemented using a rapid control prototyping platform integrated by a dSPACE board. Experimental tests are carried out for a topology with and without AES to illustrate the improvement of the proposed EV architecture.

Real‐time model‐free resilient control for discrete nonlinear systems

AbstractMotivated by the increasing complexity of systems to be controlled and different system components that cause uncertainties, threats, and disturbances, among other system stressing phenomena. Resilient control is an important design paradigm that has attracted attention from academics, practitioners, and the industrial sector. Therefore, this paper proposes the design of a model‐free resilient control for unknown discrete‐time nonlinear systems based on a recurrent high order neural network trained with an on‐line extended Kalman filter‐based algorithm for output trajectory tracking in the presence of uncertainties, disturbances, and unmodeled dynamics. This paper also includes the stability proof of the entire proposed scheme; its applicability is shown via simulation and experimental results including a comparative analysis of the proposed controller against well‐known controllers for a three‐phase induction motor.

Design and Implementation of a Real Time Control System for a 2DOF Robot Based on Recurrent High Order Neural Network Using a Hardware in the Loop Architecture

In this paper, a real-time implementation of a sliding-mode control (SMC) in a hardware-in-loop architecture is presented for a robot with two degrees of freedom (2DOF). It is based on a discrete-time recurrent neural identification method, as well as the high performance obtained from the advantages of this architecture. The identification process uses a discrete-time recurrent high-order neural network (RHONN) trained with a modified extended Kalman filter (EKF) algorithm. This is a method for calculating the covariance matrices in the EKF algorithm, using a dynamic model with the associated and measurement noises, and it increases the performance of the proposed methodology. On the other hand, the decentralized discrete-time SMC technique is used to minimize the motion error. A Virtex 7 field programmable gate array (FPGA) is configured based on a hardware-in-loop real-time implementation to validate the proposed controller. A series of several experiments demonstrates the robustness of the algorithm, as well as its immunity to noise and the inherent robustness to external perturbation, as are typically found in the input reference signals of a 2DOF manipulator robot.

Inducting Chaos on a Drone Network

Abstract This paper presents a trajectory tracking method for the formation flight of drones, which based on the synchronization of desired references in trajectory by means of complex networks in discrete-time. The pinning sliding modes control is used to track the desired trajectory controlling only the trajectory of the leader drone, and the coupling strengths are updated online during the process to achieve a smooth transition towards synchronization. The whole formation is realized by means of distancing and orientation of the drones with respect to the leader drone. The results are illustrated via simulations where each node corresponds with a quadrotor modelled by means of a high-order recurrent neural network locally controlled, whose goal is achieving that the whole drone network track an attractor chaotic type Rossler.

Real‐time neural observer‐based controller for unknown nonlinear discrete delayed systems

SummaryThis work presents a neural observer‐based controller for uncertain nonlinear discrete‐time systems with unknown time‐delays. The proposed neural observer does not need previous knowledge of the model about the system under consideration, neither the value of its parameters, delays, nor their explicit estimations. The proposed neural observer is based on a neural network composed of two recurrent high order neural networks (RHONNs) for nonmeasurable state variables, one in a parallel configuration, and for measurable state variables one in a series‐parallel configuration. The neural network is trained on‐line with an extended Kalman filter algorithm. The proposed RHONN observer provides a mathematical model for the system. Based on such a resulting mathematical model, a control law is designed using discrete‐time sliding mode block control. Applicability is presented using real‐time results that show the performance of the proposal using a linear induction motor prototype as the selected system; this prototype is under the presence of varying time‐delays. A Lyapunov analysis is included to prove the semi‐globally uniformly ultimately boundedness (SGUUB) of the proposed RHONN observer‐controller scheme for uncertain nonlinear discrete‐time systems with unknown delays.

Neural-Impulsive Pinning Control for Complex Networks Based on V-Stability

In this work, a neural impulsive pinning controller for a twenty-node dynamical discrete complex network is presented. The node dynamics of the network are all different types of discrete versions of chaotic attractors of three dimensions. Using the V-stability method, we propose a criterion for selecting nodes to design pinning control, in which only a small fraction of the nodes is locally controlled in order to stabilize the network states at zero. A discrete recurrent high order neural network (RHONN) trained with extended Kalman filter (EKF) is used to identify the dynamics of controlled nodes and synthesize the control law.

Real-time neural control of all-terrain tracked robots with unknown dynamics and network communication delays

This work focuses on the design of an intelligent controller that is a considerably large challenge for cyber-physical systems. The proposed controller can deal with unknown dynamics, actuator saturation, unknown external and internal disturbances, unknown communication delays and packet losses. Such a controller is designed using a discrete-time approach based on inverse optimal control and a recurrent high-order neural network identifier. The applicability of the proposed scheme is shown through real-time results using a tracked robot platform controlled through a wireless network under different network scenarios.

Real-time Neural Inverse Optimal Control for Low-Voltage Rid-Through enhancement of Double Fed Induction Generator based Wind Turbines

The Low-Voltage Ride-Through (LVRT) capacity of the Doubly Fed Induction Generator (DFIG) is one of the important requirements to ensure power systems stability, incorporating wind energy. While traditional control schemes present inappropriate performances under disturbances, this paper introduces a novel Neural Inverse Optimal Control (N-IOC) scheme for LVRT capacity enhancing. The developed controller is synthesized using recurrent high order neural network, which is utilized to build-up the DFIG and the DC-link dynamics. Based on such identifier, the proposed N-IOC is synthesized. This controller is experimentally validated on 1∕4 HP DFIG prototype considering various grid disturbances. Results illustrate the proposed controller effectiveness for LVRT enhancement without required decomposition process and/or any additional device.

An Autonomous Path Controller in a System on Chip for Shrimp Robot

This paper presents a path planning and trajectory tracking system for a BlueBotics Shrimp III®, which is an articulate mobile robot for rough terrain navigation. The system includes a decentralized neural inverse optimal controller, an inverse kinematic model, and a path-planning algorithm. The motor control is obtained based on a discrete-time recurrent high order neural network trained with an extended Kalman filter, and an inverse optimal controller designed without solving the Hamilton Jacobi Bellman equation. To operate the whole system in a real-time application, a Xilinx Zynq® System on Chip (SoC) is used. This implementation allows for a good performance and fast calculations in real-time, in a way that the robot can explore and navigate autonomously in unstructured environments. Therefore, this paper presents the design and implementation of a real-time system for robot navigation that integrates, in a Xilinx Zynq® System on Chip, algorithms of neural control, image processing, path planning, and inverse kinematics and trajectory tracking.

High-Order Sliding Modes Based On-Line Training Algorithm for Recurrent High-Order Neural Networks

Abstract This work presents a discrete on-line training algorithm for recurrent high-order neural networks (RHONN). The proposed training algorithm is based on the arbitrary order differentiators of high-order sliding modes (HOSM) theory. Due to HOSM-based differentiators can approximate derivatives in finite time, the proposed training algorithm avoids the compute of the derivatives, unlike conventional training algorithms. The proposed HOSM-based algorithm is implemented for the training of a RHONN identifier, and its performance is compared with the results using the extended Kalman filter (EKF) training algorithm. Results of a implementation of the identifier for the Lorenz system and an implementation of the identifier for a tracked robot using experimental data are presented.

Secondary Control of Microgrids via Neural Inverse Optimal Distributed Cooperative Control

In this paper, a new control scheme of secondary voltage and frequency control based on a discrete-time neural inverse optimal distributed cooperative structure is proposed for islanded microgrids. A neural adaptive secondary controller for each distributed generator is developed to achieve the desired goals. The proposed controllers are composed of an on-line neural identification scheme on the basis of a recurrent high-order neural network using the extended Kalman filter and a nonlinear control strategy. Additionally, the proposed control scheme does not require information of all installed distributed generators neither a distributed generator model, which improves reliability. The proposed controllers are validated through simulation for an islanded AC microgrid.

Neural sliding-mode pinning control for output synchronization for uncertain general complex networks

A novel approach for output synchronization for uncertain general complex networks with non-identical nodes is proposed. This goal can be solved applying to a small fraction of network nodes (pinned ones) a neural controller; this controller is composed of an on-line identifier based on a recurrent high-order neural network, and a sliding-mode controller. An illustrative example is included, which is composed of a network of ten nodes, with different self-dynamics, illustrating the effectiveness and good performance of the proposed control scheme.

Improving Microgrid Low-Voltage Ride-Through Capacity Using Neural Control

In this article, a neural sliding-mode linearization controller is proposed to regulate the generated active and reactive power for each distributed energy resource in a microgrid. The developed controller is based on recurrent high-order neural network identification, trained online with an extended Kalman filter learning algorithm. Based on such neural identification, adequate models of the microgrid generation units are obtained even in the presence of grid disturbances, which helps the proposed controller to reject disturbances, to ensure stability, and to operate the renewable energy sources under different grid scenarios. The proposed microgrid is composed of a wind power system, a solar power system, a battery bank, and a load demand. In addition, the microgrid under study is interconnected to an IEEE nine-bus system. The whole system is simulated in real time using the Opal-RT (OP5600) simulator. Real-time simulation results illustrate the effectiveness of the proposed control scheme to achieve trajectory tracking of the distributed energy resources active and reactive power even in the presence of grid disturbances.

Exact Takagi-Sugeno descriptor models of recurrent high-order neural networks for control applications

This work presents an exact Takagi-Sugeno descriptor model of a recurrent high-order neural network arising from identification of a nonlinear plant. The proposed rearrangement allows exploiting the nonlinear characteristics of the neural model for $$\mathcal H_\infty $$-optimal controller design whose conditions are expressed as linear matrix inequalities. Simulation and real-time results are presented that illustrate the advantages of the proposal.

A Novel Identification-Based Convex Control Scheme via Recurrent High-Order Neural Networks: An Application to the Internal Combustion Engine

This paper proposes an identification-based nonlinear control scheme which casts the plant model as a recurrent high-order neural network. The model thus obtained consists on polynomials of a fixed number of nonlinearities, a fact that is exploited by transforming it into an exact tensor-product representation whose nested convex sums may increase with the network order while preserving the number of different interpolating functions. Convexity is then used along the direct Lyapunov method to find conditions for controller design in the form of linear matrix inequalities or sum-of-squares; thanks to the fixed number of nonlinearities, they can be made progressively more relaxed while preventing the computational burden usually associated with Polya-like relaxations. The control law thus obtained is a generalization of the well-known parallel distributed compensation; its effectiveness is illustrated in academic examples and an internal combustion engine setup.

Real-Time Neural Inverse Optimal Control for a Wind Generator

This paper presents a discrete-time inverse optimal control scheme using a neural network for a doubly fed induction generator (DFIG). The DFIG generation scheme has a voltage-source converter connected between the rotor and the electrical grid, which is composed by two insulated-gate bipolar transistors (IGBT) power converters connected in back-to-back configuration. These converters are known as rotor side converter (RSC) and grid side converter (GSC), respectively. The RSC is used to control the electric torque and reactive power of the DFIG, and the GSC is used to control the dc-link voltage in the IGBT connection and the stator terminals reactive power. To take into account possible parameter variations, a recurrent high order neural network (RHONN) is used to approximate the DFIG model in an identification process; after that, based on the neural model obtained, a discrete-time inverse optimal control scheme for the RSC is developed. Using a similar approach, a dc-link neural identifier and a controller are proposed for the GSC. The proposed control scheme applicability is validated by means simulations, which includes a comparison with proportional-integral controllers; then, this control scheme is implemented in real time. The paper novelty consists on the synthesis and real-time implementation of a DFIG inverse optimal control based on a RHONN.

Neural Inverse Optimal Control Implementation for Induction Motors via Rapid Control Prototyping

This paper presents a discrete-time neural inverse optimal control for induction motors, which is implemented on a rapid control prototyping (RCP) system using a C2000 Microcontroller-Simulink platform. Such controller addresses the solution of three issues: system identification, trajectory tracking, and state estimation, which are solved independently. The neural controller is based on a recurrent high order neural network (RHONN), which is trained with an extended Kalman filter. The RHONN is an identifier to obtain an accurate motor model, which is robust to external disturbances and parameter variations. The inverse optimal controller is used to force the system to track a desired trajectory and to reject undesired disturbances. Moreover, the controller is based on a neural model and does not need the a-priori knowledge of motor parameters. A supertwisting observer is implemented to estimate the rotor magnetic fluxes. The hub of the RCP system is a TMS320f28069M MCU, which is an embedded combination of a 32-bit C28x DSP core and a real-time control accelerator. This Microcontroller is fully programmable from the Simulink environment. Simulation and experimental results illustrate the performance of the proposed controller and the RCP system, and a comparison with a control algorithm without the neural identifier is also included.

Trajectory Tracking on Uncertain Complex Networks via NN-Based Inverse Optimal Pinning Control.

A new approach for trajectory tracking on uncertain complex networks is proposed. To achieve this goal, a neural controller is applied to a small fraction of nodes (pinned ones). Such controller is composed of an on-line identifier based on a recurrent high-order neural network, and an inverse optimal controller to track the desired trajectory; a complete stability analysis is also included. In order to verify the applicability and good performance of the proposed control scheme, a representative example is simulated, which consists of a complex network with each node described by a chaotic Lorenz oscillator.

A matrix decomposition based adaptive control scheme for a class of MIMO non-canonical approximation systems

This paper presents a unified study on adaptive control of multi-input multi-output (MIMO) uncertain approximation systems in non-canonical forms. For adaptive control of such systems, this paper develops a new state feedback parametrization-based adaptive control scheme, by specifying relative degrees and establishing a normal form, deriving new system reparametrizations needed for adaptive control designs, and constructing adaptive state feedback parametrized controllers for the controlled plant with different vector relative degrees, where the uncertain control gain matrices are handled using a matrix decomposition technique. The main features of such a new control framework are demonstrated by new results for non-canonical MIMO recurrent high-order neural network systems with closed-loop stability and output tracking performance analyzed. An illustrative example with simulation results is presented to show the control design procedure and desired system performance.

High-Order Neural Networks are Equivalent to Ordinary Neural Networks

Equivalence of computational systems can assist in obtaining abstract systems, and thus enable better understanding of issues related their design and performance. For more than four decades, artificial neural networks have been used in many scientific applications to solve classification problems as well as other problems. Since the time of their introduction, multilayer feedforward neural network referred as Ordinary Neural Network (ONN), that contains only summation activation (Sigma) neurons, and multilayer feedforward High-order Neural Network (HONN), that contains Sigma neurons, and product activation (Pi) neurons, have been treated in the literature as different entities. In this work, we studied whether HONNs are mathematically equivalent to ONNs. We have proved that every HONN could be converted to some equivalent ONN. In most cases, one just needs to modify the neuronal transfer function of the Pi neuron to convert it to a Sigma neuron. The theorems that we have derived clearly show that the original HONN and its corresponding equivalent ONN would give exactly the same output, which means; they can both be used to perform exactly the same functionality. We also derived equivalence theorems for several other non-standard neural networks, for example, recurrent HONNs and HONNs with translated multiplicative neurons. This work rejects the hypothesis that HONNs and ONNs are different entities, a conclusion that might initiate a new research frontier in artificial neural network research.