Convergence theory for multi‐input discrete‐time iterative learning control with Coulomb friction,continuous outputs,and input bounds

In this paper we consider the problem of discrete‐time iterative learning control (ILC) for position trajectory tracking of multiple‐input, multiple‐output systems with Coulomb friction, bounds on the inputs, and equal static and sliding coefficients of friction. We present an ILC controller and a proof of convergence to zero tracking error, provided the associated learning gain matrices are scalar‐scaled with a sufficiently small positive scalar. We also show that non‐diagonal learning gain matrices satisfying the same prescribed conditions do not lead to the same convergence property. To the best of our knowledge, for problems with Coulomb friction, this paper represents a first convergence theory for the discrete‐time ILC problem with multiple‐bounded‐inputs and multiple‐outputs; previous work presented theory only for the single‐input, single‐output problem. Copyright © 2004 John Wiley & Sons, Ltd.     

Iterative learning control for robotic manipulators: A bounded‐error algorithm

This paper presents a model‐based nonlinear iterative learning control (NILC) for nonlinear multiple‐input and multiple‐output mechanical systems of robotic manipulators. An algorithm of a new strategy for the NILC implementation is proposed. This algorithm ensures that trajectory‐tracking errors of the proposed NILC, when implemented, are bounded by a given error norm bound. Both standard and bounded‐error learning control laws with feedback controllers attached are considered. The NILC synthesis is based on a dynamic model of a six degrees of freedom robotic manipulator. The dynamic model includes viscous and Coulomb friction and input generalized torques are bounded. With respect to the bounded‐error and standard learning processes applied to a virtual PUMA 560 robot (Unimation, Inc. Danburry, CT, USA), simulation results are presented in order to verify maximal tracking errors, convergence and applicability of the proposed learning control. Copyright © 2013 John Wiley & Sons, Ltd.     

Learning‐based iterative modular adaptive control for nonlinear systems

In this paper, we study the problem of adaptive trajectory tracking control for a class of nonlinear systems with structured parametric uncertainties. We propose to use an iterative modular approach: we first design a robust nonlinear state feedback that renders the closed‐loop input‐to‐state stable (ISS). Here, the input is considered to be the estimation error of the uncertain parameters, and the state is considered to be the closed‐loop output tracking error. Next, we propose an iterative adaptive algorithm, where we augment this robust ISS controller with an iterative data‐driven learning algorithm to estimate online the parametric uncertainties of the model. We implement this method with two different learning approaches. The first one is a data‐driven multiparametric extremum seeking method, which guarantees local convergence results, and the second is a Bayesian optimization‐based method called Gaussian Process Upper Confidence Bound, which guarantees global results in a compact search set. The combination of the ISS feedback and the data‐driven learning algorithms gives a learning‐based modular indirect adaptive controller. We show the efficiency of this approach on a two‐link robot manipulator numerical example.     

Fault‐tolerant control allocation for Mars entry vehicle using adaptive control

Accurate and reliable control of planetary entry is a major challenge for planetary exploration vehicles. For Mars entry, uncertainties in atmospheric properties such as winds aloft and density pose a major problem for meeting precision landing requirements. Anticipated manned missions to Mars will also require levels of safety and fault tolerance not required during earlier robotic missions. This paper develops a nonlinear fault‐tolerant controller specifically tailored for addressing the unique environmental and mission demands of future Mars entry vehicles. The controller tracks a desired trajectory from entry interface to parachute deployment, and has an adaptation mechanism that reduces tracking errors in the presence of uncertain parameters such as atmospheric density, and vehicle properties such as aerodynamic coefficients and inertias. This nonlinear control law generates the commanded moments for a discrete control allocation algorithm, which then generates the optimal controls required to follow the desired trajectory. The reaction control system acts as a non‐uniform quantizer, which generates applied moments that approximate the desired moments generated by a continuous adaptive control law. If a fault is detected in the control jets, it reconfigures the controls and minimizes the impact of control failures or damage on trajectory tracking. It is assumed that a fault identification and isolation scheme already exists to identify failures. A stability analysis is presented, and fault tolerance performance is evaluated with non real‐time simulation for a complete Mars entry trajectory tracking scenario using various scenarios of control effector failures. The results presented in the paper demonstrate that the control algorithm has a satisfactory performance for tracking a pre‐defined trajectory in the presence of control failures, in addition to plant and environment uncertainties. Copyright © 2010 John Wiley & Sons, Ltd.     

Stochastic high‐order internal model‐based adaptive TILC with random uncertainties in initial states and desired reference points

Many practical batch processes operate repetitively in industry and lack intermediate measurements for the interested process variables. Moreover, the initial states as well as the desired product objective often vary with different runs because of the existence of many uncertainties in practice. This work proposes a novel adaptive terminal iterative learning control method to deal with random uncertainties in desired terminal points and initial states. The run‐varying initial states are formulated by a stochastic high‐order internal model, which is further incorporated into the controller design. The desired terminal output is run dependent and is directly compensated like a feedback term in the controller. Only the system output at the endpoint of an operation is utilized to update the control signal. An estimation algorithm is designed to update the system Markov parameters as a whole. No explicit model information is involved in the controller design; thus, the proposed method is data driven and can be applied to nonlinear systems directly. Both the theoretical analysis and the simulation studies demonstrate the effectiveness of the proposed approach under random initial states and iteration‐varying referenced terminal points. Copyright © 2016 John Wiley & Sons, Ltd.     

Stability analysis of quantized iterative learning control systems using lifting representation

This paper presents a stability analysis of the iterative learning control for discrete‐time systems with data quantization. Three quantized iterative learning control schemes are considered by using different quantized signals, including system output quantized signal, tracking error quantized signal, and control input quantized signal. The logarithmic quantizer is introduced to decode these signals with a number of quantization levels, and the sector bound method is used to deal with the quantization error. Based on the supervector formulation for iterative learning control systems, some convergence conditions for these iterative learning control laws are given, respectively. It is shown that iterative learning control laws with system output quantized signal and control input quantized signal only guarantee that the tracking error converges to a bound and the bound depending on quantization density and desired trajectory. Thus, the iterative learning control law with tracking error quantized signal can obtain zero tracking error. These results are illustrated by 2 examples.     

Novel model reference adaptive control architecture using semi‐initial excitation‐based switched parameter estimator

This paper is a generalization of the recently developed techniques of initial excitation (IE)–based adaptive control with an introduction to the definition of semi‐initial excitation (semi‐IE), a still more relaxed notion than IE. Classical adaptive controllers typically ensure Lyapunov stability of the extended error dynamics (tracking error + parameter estimation error) and asymptotic tracking, while requiring a stringent condition of persistence of excitation (PE) for parameter convergence. Of late, the authors have proposed a new adaptive control architecture, which guarantees parameter convergence under the online‐verifiable IE condition leading to exponential stability of the extended error dynamics. In earlier works, it has been established that the IE condition is significantly milder than the classical PE condition. The current work further slackens the excitation condition by proposing the concept of semi‐IE. The proposed adaptive controller is proved to ensure convergence of the parameter estimation error to a lower‐dimensional manifold under the weaker semi‐IE condition, while the stronger condition of IE guarantees convergence of the parameter estimation error to zero. The designed algorithm is shown to improve transient response of tracking error sufficiently in contrast to conventional adaptive controllers.     

Output‐feedback H∞ quadratic tracking control of linear systems using reinforcement learning

This paper presents an online learning algorithm based on integral reinforcement learning (IRL) to design an output‐feedback (OPFB) H_∞ tracking controller for partially unknown linear continuous‐time systems. Although reinforcement learning techniques have been successfully applied to find optimal state‐feedback controllers, in most control applications, it is not practical to measure the full system states. Therefore, it is desired to design OPFB controllers. To this end, a general bounded L₂ ‐gain tracking problem with a discounted performance function is used for the OPFB H_∞ tracking. A tracking game algebraic Riccati equation is then developed that gives a Nash equilibrium solution to the associated min‐max optimization problem. An IRL algorithm is then developed to solve the game algebraic Riccati equation online without requiring complete knowledge of the system dynamics. The proposed IRL‐based algorithm solves an IRL Bellman equation in each iteration online in real time to evaluate an OPFB policy and updates the OPFB gain using the information given by the evaluated policy. An adaptive observer is used to provide the knowledge of the full states for the IRL Bellman equation during learning. However, the observer is not needed after the learning process is finished. A simulation example is provided to verify the convergence of the proposed algorithm to a suboptimal OPFB solution and the performance of the proposed method.     

Robust adaptive control for non‐linear systems in generalized output‐feedback canonical form

Design of global robust adaptive output‐feedback dynamic compensators for stabilization and tracking of a class of systems that are globally diffeomorphic into systems in generalized output‐feedback canonical form is investigated. This form includes as special cases the standard output‐feedback canonical form and various other forms considered previously in the literature. Output‐dependent non‐linearities are allowed to enter both additively and multiplicatively. The system is allowed to contain unknown parameters multiplying output‐dependent non‐linearities and, also, unknown non‐linearities satisfying certain bounds. Under the assumption that a constant matrix can be found to achieve a certain property, it is shown that a reduced‐order observer and a backstepping controller can be designed to achieve practical stabilization of the tracking error. If this assumption is not satisfied, it is shown that the control objective can be achieved by introducing additional dynamics in the observer. Sufficient conditions under which asymptotic tracking and stabilization can be achieved are also given. This represents the first robust adaptive output‐feedback tracking results for this class of systems. Copyright © 2003 John Wiley & Sons, Ltd.     

Iterative learning control for non‐repetitive trajectory tracking of robot manipulators with joint position constraints and actuator faults

In this work, we present a novel iterative learning control (ILC) scheme for a class of joint position constrained robot manipulator systems with both multiplicative and additive actuator faults. Unlike most ILC literature that requires identical reference trajectory from trail to trail, in this work the reference trajectory can be non‐repetitive over the iteration domain without assuming the identical initial condition. A tan‐type Barrier Lyapunov Function is proposed to deal with the constraint requirements which can be both time and iteration varying, with ILC update laws adopted to learn the iteration‐invariant system uncertainties, and robust methods used to compensate the iteration and time varying actuator faults and disturbances. We show that under the proposed ILC scheme, uniform convergence of the full state tracking error beyond a small time interval in each iteration can be guaranteed over the iteration domain, while the constraint requirements on the joint position vector will not be violated during operation. An illustrative example on a two degree‐of‐freedom robotic manipulator is presented to demonstrate the effectiveness of the proposed control scheme. Copyright © 2016 John Wiley & Sons, Ltd.     

基于神经网络的学习控制及其在机器人中的应用

针对一类非线性系统的跟踪控制问题 ,首先提出了一种遗忘因子迭代学习控制算法 ,给出了算法收敛的充分条件 ,然后 ,利用神经网络原理 ,对要求跟踪的新的期望轨迹 ,在系统的历史控制经验基础上 ,用神经网络估计系统的期望控制输入 ,然后将其作为迭代学习控制器的初始控制输入 ,再由迭代学习律逐步改善控制输入 ,使系统的实际输出只需较少的迭代次数就能达到跟踪的精度要求。机器人系统的仿真结果表明了该算法的有效性。     

Experience replay–based output feedback Q‐learning scheme for optimal output tracking control of discrete‐time linear systems

This paper focuses on solving the adaptive optimal tracking control problem for discrete‐time linear systems with unknown system dynamics using output feedback. A Q‐learning‐based optimal adaptive control scheme is presented to learn the feedback and feedforward control parameters of the optimal tracking control law. The optimal feedback parameters are learned using the proposed output feedback Q‐learning Bellman equation, whereas the estimation of the optimal feedforward control parameters is achieved using an adaptive algorithm that guarantees convergence to zero of the tracking error. The proposed method has the advantage that it is not affected by the exploration noise bias problem and does not require a discounting factor, relieving the two bottlenecks in the past works in achieving stability guarantee and optimal asymptotic tracking. Furthermore, the proposed scheme employs the experience replay technique for data‐driven learning, which is data efficient and relaxes the persistence of excitation requirement in learning the feedback control parameters. It is shown that the learned feedback control parameters converge to the optimal solution of the Riccati equation and the feedforward control parameters converge to the solution of the Sylvester equation. Simulation studies on two practical systems have been carried out to show the effectiveness of the proposed scheme.     

Adaptive robust control of a class of nonlinear systems in semi‐strict feedback form with non‐uniformly detectable unmeasured internal states

This paper proposes a novel control method for a special class of nonlinear systems in semi‐strict feedback form. The main characteristic of this class of systems is that the unmeasured internal states are non‐uniformly detectable, which means that no observer for these states can be designed to make the observation error exponentially converge to zero. In view of this, a projection‐based adaptive robust control law is developed in this paper for this kind of system. This method uses a projection‐type adaptation algorithm for the estimation of both the unknown parameters and the internal states. Robust feedback term is synthesized to make the system robust to uncertain nonlinearities and disturbances. Although the estimation error for both the unknown parameters and the internal states may not converge to zero, the tracking error of the closed‐loop system is proved to converge to zero asymptotically if the system has only parametric uncertainties. Furthermore, it is theoretically proved that all the signals are bounded, and the control algorithm is robust to bounded disturbances and uncertain nonlinearities with guaranteed output tracking transient performance and steady‐state accuracy in general. The class of system considered here has wide engineering applications, and a practical example—control of mechanical systems with dynamic friction—is used as a case study. Simulation results are obtained to demonstrate the applicability of the proposed control methodology. Copyright © 2010 John Wiley & Sons, Ltd.     

Neural network adaptive robust control with application to precision motion control of linear motors

In this paper, neural networks (NNs) and adaptive robust control (ARC) design philosophy are integrated to design performance‐oriented control laws for a class of single‐input–single‐output (SISO) nth‐order non‐ linear systems. Both repeatable (or state dependent) unknown non‐linearities and non‐repeatable unknown non‐linearities such as external disturbances are considered. In addition, unknown non‐linearities can exist in the control input channel as well. All unknown but repeatable non‐linear functions are approximated by outputs of multi‐layer neural networks to achieve a better model compensation for an improved performance. All NN weights are tuned on‐line with no prior training needed. In order to avoid the possible divergence of the on‐line tuning of neural network, discontinuous projection method with fictitious bounds is used in the NN weight adjusting laws to make sure that all NN weights are tuned within a prescribed range. By doing so, even in the presence of approximation error and non‐repeatable non‐linearities such as disturbances, a controlled learning is achieved and the possible destabilizing effect of on‐line tuning of NN weights is avoided. Certain robust control terms are constructed to attenuate various model uncertainties effectively for a guaranteed output tracking transient performance and a guaranteed final tracking accuracy in general. In addition, if the unknown repeatable model uncertainties are in the functional range of the neural networks and the ideal weights fall within the prescribed range, asymptotic output tracking is also achieved to retain the perfect learning capability of neural networks in the ideal situation. The proposed neural network adaptive control (NNARC) strategy is then applied to the precision motion control of a linear motor drive system to help to realize the high‐performance potential of such a drive technology. NN is employed to compensate for the effects of the lumped unknown non‐linearities due to the position dependent friction and electro‐magnetic ripple forces. Comparative experiments verify the high‐performance nature of the proposed NNARC. With an encoder resolution of 1 µm, for a low‐speed back‐and‐forth movement, the position tracking error is kept within ±2 µm during the most execution time while the maximum tracking error during the entire run is kept within ±5.6 µm. Copyright © 2001 John Wiley & Sons, Ltd.     

Model and experience‐based initial input construction for iterative learning control

The initial choice of input in iterative learning control (ILC) generally has a significant effect on the error incurred over subsequent trials. In this paper, techniques are developed that use experimental data gathered over previous applications of ILC in order to generate an initial input signal for the tracking of a new reference trajectory. A model‐based approach is then incorporated to overcome the limitation of insufficient previous experimental data, and a robust design procedure is developed. Experimental evaluation results are obtained using a gantry robot facility. Copyright © 2010 John Wiley & Sons, Ltd.     

Output‐feedback model‐reference adaptive calibration for map‐based anti‐jerk control of electromechanical automotive clutches

It is well known that the map‐based control can reduce the computational burden of the automotive on‐board controller. This paper proposes an output‐feedback model‐reference adaptive control algorithm to calibrate the map‐based anti‐jerk controller for electromechanical clutch engagement. The algorithm can be used to adaptively construct a data‐driven fuzzy rule base without resorting to manual tuning, so that it can overcome the problem of conventional knowledge‐based fuzzy logic design, which involves strenuous parameter‐tuning work in the construction of calibration maps. To accurately define the consequent of each fuzzy rule for anti‐jerk control, an output feedback law for computing the reference trajectory of clutch engagement is developed to eliminate the discontinuous slip‐stick transition, whereas an adaptive controller is designed to track the reference trajectory and compensate the nonlinearity. The convergence of the proposed output‐feedback model‐reference adaptive control algorithm is analyzed. Simulation results indicate that the proposed method can successfully reduce the excessive vehicle jerk and frictional energy dissipation during clutch engagement as compared with the conventional knowledge‐based fuzzy logic controller without fine tuning.     

Model reference adaptive iterative learning control for linear systems

In this paper, we propose a model reference adaptive control (MRAC) strategy for continuous‐time single‐input single‐output (SISO) linear time‐invariant (LTI) systems with unknown parameters, performing repetitive tasks. This is achieved through the introduction of a discrete‐type parametric adaptation law in the ‘iteration domain’, which is directly obtained from the continuous‐time parametric adaptation law used in standard MRAC schemes. In fact, at the first iteration, we apply a standard MRAC to the system under consideration, while for the subsequent iterations, the parameters are appropriately updated along the iteration‐axis, in order to enhance the tracking performance from iteration to iteration. This approach is referred to as the model reference adaptive iterative learning control (MRAILC). In the case of systems with relative degree one, we obtain a pointwise convergence of the tracking error to zero, over the whole finite time interval, when the number of iterations tends to infinity. In the general case, i.e. systems with arbitrary relative degree, we show that the tracking error converges to a prescribed small domain around zero, over the whole finite time interval, when the number of iterations tends to infinity. It is worth noting that this approach allows: (1) to extend existing MRAC schemes, in a straightforward manner, to repetitive systems; (2) to avoid the use of the output time derivatives, which are generally required in traditional iterative learning control (ILC) strategies dealing with systems with high relative degree; (3) to handle systems with multiple tracking objectives (i.e. the desired trajectory can be iteration‐varying). Finally, simulation results are carried out to support the theoretical development. Copyright © 2006 John Wiley & Sons, Ltd.     

Iterative learning control for a class of non‐affine‐in‐input processes in Hilbert space

In this paper, iterative learning control (ILC) of a class of non‐affine‐in‐input processes is considered in Hilbert space, where the plant operators are quite general in the sense that they could be static or dynamic, differentiable or non‐differentiable, continuous‐time or discrete‐time, and so forth. The control problem is first transformed to a problem of solving global implicit function to ensure the uniqueness of desired control input. Then, two contraction mapping‐based ILC schemes are proposed in terms of the continuous differentiability of process model, where the learning convergence condition is derived through rigorous analysis. The proposed ILC schemes make full use of the process repetition, deal with system uncertainties easily, and are effective to infinite‐dimensional or distributed parameter systems. In the end, the learning controller is applied to the boundary output control of a class of anaerobic digestion process for wastewater treatment. The control efficacy is verified by simulation. Copyright © 2013 John Wiley & Sons, Ltd.     

Modeling of non‐linear systems by FWNNs and their intelligent control

By utilizing some of the important properties of wavelets like denoising, compression, multiresolution along with the concepts of fuzzy logic and neural network, two fuzzy wavelet neural networks (FWNNs) are proposed for approximating any arbitrary non‐linear function, hence, identifying a non‐linear system. We have fuzzified the output of DWT block, which receives the given inputs, in the proposed two methods: one using compression property and other using multiresolution property. We present a new type of fuzzy neuron model, each non‐linear synapse of which is characterized by a set of fuzzy implication rules with singleton weights in their consequents. It is shown that noise and disturbance in the reference signal are reduced with wavelets and also the variation of somatic gain, the parameter that controls the slope of the activation function in the neural network, leads to more accurate output. Identification results are found to be accurate and speed of their convergence is fast. Next, we simulate a control system for keeping output at a desired level by using the identified models. Two self‐learning controllers are designed in this simulation. One is a self‐learning fuzzy PI controller and other is a NN controller. Simulation results show that the NN controller is more adaptive and fast. Copyright © 2005 John Wiley & Sons, Ltd.     

Neural‐hybrid control of systems with sandwiched dead‐zones

The control of systems that have sandwiched nonsmooth nonlinearities, such as a dead‐zone sandwiched between two dynamic blocks, is addressed. An adaptive inverse control scheme using a hybrid controller structure and a neural network based inverse compensator, is proposed for such systems with unknown sandwiched dead‐zone. This neural‐hybrid controller consists of an inner loop discrete‐time feedback structure incorporated with an adaptive inverse using a neural network for the unknown dead‐zone, and an outer‐loop continuous‐time feedback control law for achieving desired output tracking. The dead‐zone compensator consists of two neural networks, one used as an estimator of the sandwiched dead‐zone function and the other for the compensation itself. The compensator neural network has neurons that can approximate jump functions such as a dead‐zone inverse. The weights of the two neural networks are tuned using a modified gradient algorithm. Simulation results are given to illustrate the performance of the proposed neural‐hybrid controller. Copyright © 2002 John Wiley & Sons, Ltd.