A critical review of the most popular types of neuro control

In this review article, the most popular types of neural network control systems are briefly introduced and their main features are reviewed. Neuro control systems are defined as control systems in which at least one artificial neural network (ANN) is directly involved in generating the control command. Initially, neural networks were mostly used to model system dynamics inversely to produce a control command which pushes the system towards a desired or reference value of the output (1989). At the next stage, neural networks were trained to track a reference model, and ANN model reference control appeared (1990). In that method, ANNs were used to extend the application of adaptive reference model control, which was a well‐known control technique. This attitude towards the extension of the application of well‐known control methods using ANNs was followed by the development of ANN model‐predictive (1991), ANN sliding mode (1994) and ANN feedback linearization (1995) techniques. As the first category of neuro controllers, inverse dynamics ANN controllers were frequently used to form a control system together with other controllers, but this attitude faded as other types of ANN control systems were developed. However, recently, this approach has been revived. In the last decade, control system designers started to use ANNs to compensate/cancel undesired or uncertain parts of systems' dynamics to facilitate the use of well‐known conventional control systems. The resultant control system usually includes two or three controllers. In this paper, applications of different ANN control systems are also addressed. Copyright © 2011 John Wiley and Sons Asia Pte Ltd and Chinese Automatic Control Society     

Design and analysis of gain-scheduled control using local controller networks

This paper describes a nonlinear control structure known as a local controller network. The structure consists of a weighted combination of a number of individual controllers, each of which is valid locally in the state space of the plant. Local controller designs are based upon local models valid in operating regimes which do not necessarily contain any physical equilibria. Consequently, the transient performance can be improved. Some 'scheduling' variables determine the current operating regime, and a validity function is assigned to each local controller. A 'feedforward' component may be used in each local controller in order to compensate directly for the operating-point-dependent model offsets. The application of the local controller network approach to a nonlinear control problem, that of longitudinal vehicle dynamics control, is described. A stability analysis for the discrete-time local controller network is given in this paper and the results are compared with known theoretical guidelines for related control approaches such as gain scheduling and feedback linearization.     

An analytical comparison of a neural network and a model-basedadaptive controller

A neural network inverse dynamics controller with adjustable weights is compared with a computed-torque type adaptive controller. Lyapunov stability techniques, usually applied to adaptive systems, are used to derive a globally asymptotically stable adaptation law for a single-layer neural network controller that bears similarities to the well-known delta rule for neural networks. This alternative learning rule allows the learning rates of each connection weight to be individually adjusted to give faster convergence. The role of persistently exciting inputs in ensuring parameter convergence, often mentioned in the context of adaptive systems, is emphasized in relation to the convergence of neural network weights. A coupled, compound pendulum system is used to develop inverse dynamics controllers based on adaptive and neural network techniques. Adaptation performance is compared for a model-based adaptive controller and a simple neural network utilizing both delta-rule learning and the alternative adaptation law.     

Feedback linearization based computer controlled medication design for automatic treatment of parturient paresis of cows

Based on an existing model for calcium homeostatis (dynamics) and taking the help of feedback linearization philosophy of nonlinear control theory, two control design (medication) strategies are presented for automatic treatment of parturient paresis (milk fever) disease of cows. An important advantage of the new approach is that it results in a simple and straightforward method and eliminates the necessity of a significantly more complex neural network based nonlinear optimal control technique, as proposed by the author earlier. As an added advantage, unlike the neural network technique, the new approach leads to 'closed form solution' for the nonlinear controller. Moreover, global asymptotic stability of the closed loop system is always guaranteed. Besides theoretical justifications, the resulting controllers (medication strategies) are validated from numerical simulation studies of the nonlinear system as well. Moreover, from a numerical study about the robustness of the algorithms with respect to parametric uncertainty, it was observed that the optimal control formulation is a better option over the dynamic inversion formulation.     

Non-linear Control of Heave for an Unmanned Helicopter Using a Neural Network

This paper describes a new non-linear control technique applied to the heave control of an unmanned rotorcraft. First a hybrid plant model consisting of exactly known dynamics is combined with a black-box representation of the unknown dynamics. Desired trajectories are calculated to smoothly achieve a sequence of random step changes in desired height according to certain optimal criterion and plant limitations. Control inputs are then determined using the MATLAB^® optimisation toolbox to achieve those desired trajectories for the plant heave model. Finally, a neural network is trained to mimic the control inputs resulting from the optimisation process. The neural network controller produces trajectories closely resembling the results from the optimisation process but with a much reduced computation time. Flight test results of control of the heave dynamics of a helicopter confirm the neural network controller’s ability to operate in high disturbance conditions and outperform a proportional-derivative (PD) controller.     

Intelligent control of exit temperature in a gas-fuel can-type combustor

The dynamical characteristics of a gas-fuel can-type combustor are highly nonlinear and are too complicated to be modeled precisely. Consequently, it is very difficult to control the exit temperature in a combustor using a conventional feedback controller. This paper investigates the models, describing the dynamics of exit temperature for a gas-fuel can-type combustor, and designs the intelligent controllers, based on the characteristics of the constructed models, to control the exit temperature in the combustor. An identified neural network (INN) was utilized to construct the dynamical models because of its powerful learning and handling ability for nonlinear systems. According to the open-loop responses of the investigated models, two controllers, a self-tuning fuzzy proportional–integral–derivative controller and a neural network controller, were developed for the exit temperature control. Experiments were conducted to evaluate the constructed models and the designed controllers.     

Unified stabilizing controller synthesis approach for discrete-time intelligent systems with time delays by dynamic output feedback

A novel model, termed the standard neural network model (SNNM), is advanced to describe some delayed (or non-delayed) discrete-time intelligent systems com- posed of neural networks and Takagi and Sugeno (T-S) fuzzy models. The SNNM is composed of a discrete-time linear dynamic system and a bounded static nonlinear operator. Based on the global asymptotic stability analysis of the SNNMs, linear and nonlinear dynamic output feedback controllers are designed for the SNNMs to stabilize the closed-loop systems, respectively. The control design equations are shown to be a set of linear matrix inequalities (LMIs) which can be easily solved by various convex optimization algorithms to determine the control signals. Most neural-network-based (or fuzzy) discrete-time intelligent systems with time delays or without time delays can be transformed into the SNNMs for controller synthesis in a unified way. Three application examples show that the SNNMs not only make controller synthesis of neural-network-based (or fuzzy) discrete-time intelligent systems much easier, but also provide a new approach to the synthesis of the controllers for the other type of nonlinear systems.     

Symbolic control of linear systems based on symbolic subsystems

This paper describes an approach to the control of continuous systems through the use of symbolic models describing the system behavior only at a finite number of points in the state space. These symbolic models can be seen as abstract representations of the continuous dynamics enabling the use of algorithmic controller design methods. We identify a class of linear control systems for which the loss of information incurred by working with symbolic subsystems can be compensated by feedback. We also show how to transform symbolic controllers designed for a symbolic subsystem into controllers for the original system. The resulting controllers combine symbolic controller dynamics with continuous feedback control laws and can thus be seen as hybrid systems. Furthermore, if the symbolic controller already accounts for software/hardware requirements, the hybrid controller is guaranteed to enforce the desired specifications by construction thereby reducing the need for formal verification.     

Design and performance of an intelligent predictive controller for a six-degree-of-freedom robot using the Elman network

The aim of this paper was to propose a recurrent neural network-based predictive controller for robotic manipulators. A neural network controller for a six-joint Stanford robotic manipulator was designed using the generalized predictive control (GPC) and the Elman network. The GPC algorithm, which is a class of digital control method, requires long computational time. This is a disadvantage in real-time robot control; therefore, the Elman network controller was designed to reduce processing time by avoiding the highly mathematical and computational complexity of the GPC. The main reason for choosing the Elman network, amongst several neural network algorithms, was that the presence of feedback loops have a profound impact on the learning capability of the network. The designed neural network controller was able to recover quickly because of its significant generalization capability, which allowed it to adapt very rapidly to changes in inputs. The performance of the controller was also shown graphically using simulation software, including the dynamics and kinematics of the robot model.     

Adaptive neural model-based fault tolerant control for multi-variable processes

An adaptive neural network model-based fault tolerant control approach for unknown non-linear multi-variable dynamic systems is proposed. A multi-layer Perceptron network is used as the process model and is adapted on-line using the extended Kalman filter to learn changes in process dynamics. In this way, the adaptive model will learn the post-fault dynamics caused by actuator or component faults. Then, the inversion of the neural model is used as a controller to maintain the system stability and control performance after fault occurrence. The convergence of the model inversion control is proved using Lyapunov method. The proposed method is applied to the simulation of a two-input two-output continuous-stirred tank reactor to demonstrate the effectiveness of the approach. Several actuator and component faults are simulated on the continuously stirred tank reactor process when the system is under the proposed fault tolerant control. The results have shown a fast recovery of tracking performance and the maintained stability.     

Neural network robust control of a 3-DOF hydraulic manipulator with asymptotic tracking

Multiaxial hydraulic manipulators are complicated systems with highly nonlinear dynamics and various modeling uncertainties, which hinders the development of high-performance controller. In this paper, a neural network feedforward with a robust integral of the sign of the error (RISE) feedback is proposed for high precise tracking control of hydraulic manipulator systems. The established nonlinear model takes three-axis dynamic coupling, hydraulic actuator dynamics, and nonlinear friction effects into consideration. A radial basis function neural network (RBFNN) is synthesized to approximate the uncertain system dynamics and external disturbance, which can greatly reduce the dependence on accurate system model. In addition, a continuous RISE feedback law is judiciously integrated to deal with the residual unknown dynamics. Since the major unknown dynamics can be estimated by the RBFNN and then compensated in the feedforward design, the high-gain feedback issue in RISE feedback control will be avoided. The proposed RISE-based neural network robust controller theoretically guarantees an excellent semi-global asymptotic stability. Comparative simulation is performed on a 3-DOF hydraulic manipulator, and the obtained results verify the effectiveness of the proposed controller.     

Output feedback performance recovery in the presence of uncertainties

We propose an output-feedback tracking controller for uncertain, nonaffine, nonlinear systems. The output feedback controller results in a closed-loop system with a three-time-scale structure; an extended high-gain observer estimates unmeasured states and uncertainties in the fastest time scale and dynamic inversion is used to deal with nonaffine inputs and input uncertainties in the intermediate time scale while the plant dynamics evolves in the slowest time scale. The dynamic inversion algorithm is based on sector conditions and results in exponential convergence of the inputs. Together with the extended high-gain observer, dynamic inversion results in performance recovery of a target system. The singular perturbation method is used to analyze the stability and performance of the system and numerical simulations are used to demonstrate the effectiveness of the control design.     

Modelling and optimal controller design of networked control systems with multiple delays

In this paper we discuss the modelling and control of networked control systems (NCS) where sensors, actuators and controllers are distributed and interconnected by a common communication network. Multiple distributed communication delays as well as multiple inputs and multiple outputs (MIMO) are considered in the modelling algorithm. In addition, the asynchronous sampling mechanisms of distributed sensors are characterized to obtain the actual time delays between sensors and the controller. Due to the characteristics of a network architecture, piecewise constant plant inputs are assumed and discrete-time models of plant and controller dynamics are adopted to analyse the stability and performance of a closed-loop NCS. The analysis result is used to verify the stability and performance of an NCS without considering the impact of multiple time delays in the controller design. In addition, the proposed NCS model is used as a foundation for optimal controller design. The proposed control algorithm utilizes the information of delayed signals and improves the control performance of a control system encountering distributed communication delays. Several simulation studies are provided to verify the control performance of the proposed controller design.     

On adaptive trajectory tracking of a robot manipulator usinginversion of its neural emulator

This paper is concerned with the design of a neuro-adaptive trajectory tracking controller. The paper presents a new control scheme based on inversion of a feedforward neural model of a robot arm. The proposed control scheme requires two modules. The first module consists of an appropriate feedforward neural model of forward dynamics of the robot arm that continuously accounts for the changes in the robot dynamics. The second module implements an efficient network inversion algorithm that computes the control action by inverting the neural model. In this paper, a new extended Kalman filter (EKF) based network inversion scheme is proposed. The scheme is evaluated through comparison with two other schemes of network inversion: gradient search in input space and Lyapunov function approach. Using these three inversion schemes the proposed controller was implemented for trajectory tracking control of a two-link manipulator. Simulation results in all cases confirm the efficacy of control input prediction using network inversion. Comparison of the inversion algorithms in terms of tracking accuracy showed the superior performance of the EKF based inversion scheme over others.     

Asymptotic tracking by a reinforcement learning-based adaptive critic controller

Adaptive critic (AC) based controllers are typically discrete and/or yield a uniformly ultimately bounded stability result because of the presence of disturbances and unknown approximation errors. A continuous-time AC controller is developed that yields asymptotic tracking of a class of uncertain nonlinear systems with bounded disturbances. The proposed AC-based controller consists of two neural networks (NNs) – an action NN, also called the actor, which approximates the plant dynamics and generates appropriate control actions; and a critic NN, which evaluates the performance of the actor based on some performance index. The reinforcement signal from the critic is used to develop a composite weight tuning law for the action NN based on Lyapunov stability analysis. A recently developed robust feedback technique, robust integral of the sign of the error (RISE), is used in conjunction with the feedforward action neural network to yield a semiglobal asymptotic result. Experimental results are provided that illustrate the performance of the developed controller.     

Autonomous mobile robots navigation using RBF neural compensator

This paper presents an approach to adaptive trajectory tracking of mobile robots which combines a feedback linearization based on a nominal model and a RBF-NN adaptive dynamic compensation. For a robot with uncertain dynamic parameters, two controllers are implemented separately: a kinematics controller and an inverse dynamics controller. The uncertainty in the nominal dynamics model is compensated by a neural adaptive feedback controller. The resulting adaptive controller is efficient and robust in the sense that it succeeds to achieve a good tracking performance with a small computational effort. The analysis of the RBF-NN approximation error on the control errors is included. Finally, the performance of the control system is verified through experiments.     

Performance recovery of a class of uncertain non-affine systems with unmodelled dynamics: an indirect dynamic inversion method

This paper presents an output feedback indirect dynamic inversion (IDI) approach for a class of uncertain nonaffine systems with input unmodelled dynamics. Compared with previous approaches to achieve performance recovery, the proposed method aims at dealing with a broader class of nonaffine-in-control systems with triangular structure. An IDI state feedback law is designed first, in which less knowledge of the model plant is needed compared to earlier approximate dynamic inversion methods, thus yielding more robust performance. After that, an extended high-gain observer is designed to accomplish the task with output feedback. Finally, we prove that the designed IDI controller is equivalent to an adaptive proportional-integral (PI) controller, with respect to both time response equivalence and robustness equivalence. The conclusion implies that for the studied strict-feedback non-affine systems with unmodelled dynamics, there always exits a PI controller to stabilise the systems. The effectiveness and benefits of the designed approach are verified by three examples.     

Adaptive neural optimal control via command filter for nonlinear multi-agent systems including time-varying output constraints

In this article, an optimal command-filtered backstepping control approach is proposed for uncertain strict-feedback nonlinear multi-agent systems (MASs) including output constraints and unmodeled dynamics. One-to-one nonlinear mapping (NM) is utilized to recast constrained systems as corresponding unrestricted systems. A dynamical signal is applied to cope with unmodeled dynamics. Based on dynamic surface control (DSC), the feedforward controller is designed by introducing error compensating signals. The optimal feedback controller is produced applying adaptive dynamic programming (ADP) and integral reinforcement learning (IRL) techniques in which neural networks are utilized to approximate the relevant cost functions online with established weight updating laws. Therefore, the entire controller, including feedforward and feedback controllers, not only ensures that all signals in the closed-loop systems are cooperative semi-globally uniformly ultimately bounded (SGUUB) and the outputs maintain in the provided time-varying constraints, but also makes sure that the cost functions achieve minimization. A simulation example is presented to illustrate the feasibility of the proposed control algorithm.     

A harmonic balance approach to robust neural control of MIMO nonlinear processes applied to a distillation column

Robust stability and performance are the two most basic features of feedback control process. The harmonic balance analysis based on the describing function technique enables to analyze the stability of limit cycles arising from a closed loop control process operating over nonlinear plants. In this work a robust stability analysis based on the harmonic balance is presented and applied to a neural network controller in series with a dynamic multivariable nonlinear plant under generic Lur’e configuration. The neural controller is replaced by its sinusoidal input describing function while a linearized model is derived to represent the nonlinear plant dynamics. The uncertainty induced by the high harmonics effect for the neural controller, together with the neglected nonlinear dynamics due to plant linearization are incorporated in the robustness analysis as structured norm bounded uncertainties. Stability and robustness conditions for the neural closed loop control system are discussed using the harmonic balance equation together with the structured singular values of the uncertainty. The application to a multivariable binary distillation column under feedback neurocontrol illustrates the usefulness of the robustness approach here developed to predict the absence of limit cycles, which of course is subject to the usual restrictions of the describing function method.     

Distributed control of plantwide chemical processes

Complex process plants increasingly appear in modern chemical industry. The wide use of material recycles and heat integration (with recycle and bypass streams) profoundly alters plantwide process dynamics and further increases their complexity. The interactions between process units may lead to poor performance of decentralized control systems. On the other hand, the complexity of plantwide systems prohibits the use of centralized controllers that reply on the complex model of the entire plantwide process. This paper addresses the plantwide chemical process control problem from a network perspective. The entire chemical plant is modeled as a network of process units linked by physical mass and energy flow and controlled by controllers that communicate with each other (i.e., distributed controllers). A two-port linear time-invariant representation is proposed to describe the dynamics of each process unit and its corresponding distributed controller. A two-step plantwide linear control design approach is developed. By using the dissipativity theory, the plantwide stability and control performance is translated into the closed-loop dissipativity condition that each distributed controller has to achieve. This allows the distributed controllers to be designed independently and to operate autonomously. The proposed approach is illustrated by a case study of a process network that consists of a reactor and a distillation column.