Elsevier

Neurocomputing

Volume 145, 5 December 2014, Pages 392-401
Neurocomputing

Immunological mechanism inspired iterative learning control

https://doi.org/10.1016/j.neucom.2014.05.013Get rights and content

Abstract

In this paper, an iterative learning control method based on the recognition, response, and memory mechanism of immune system (IRRM-ILC) is proposed. According to the immunological recognition, response, and memory processes, several new components are designed and then combined with the conventional ILC. The proposed IRRM-ILC controller can mitigate the disturbances by detecting, identifying, and memorizing them. Thereby, the anti-disturbance ability of the conventional ILC is enhanced. The proposed IRRM-ILC is applied to a temperature control system of wet spinning coagulation process. Simulation results with random and repeated disturbances demonstrate that the IRRM-ILC has a better performance than the conventional ILC in terms of convergence and stability in the presence of disturbances.

Introduction

Learning is a process for an intelligent system to acquire knowledge or experience on the basis of its perception and cognition of the environment and then to act on the environment referring the knowledge/experience so as to improve its behavior performance the next time [1]. Iterative learning control (ILC) is a well-known advanced control method which utilizes some basic learning principles to achieve control of the plant. The ILC is a recursive algorithm that utilizes the tracking discrepancy/error of the system output from the desired trajectory and iteratively updates its current control command so as to generate an upgraded control command for the next operation when the system operates repetitively over a fixed finite time interval [2]. It is based on the notion that the performance of a system that executes the same task multiple times can be improved by learning from previous executions (batches, trials, iterations, passes) [3]. It is a data-driven method that uses only the plant׳s data instead of any knowledge of the plant׳s model. In many systems such as chemical processes, robot motions, and servo systems, the same tasks are carried out repeatedly over a fixed time interval. The ILC is able to gradually improve its control performance by learning from previous actions [4].

Since introduced by Arimoto et al. in 1984 [2], the ILC has drawn much attention from researchers and has been successfully applied in practices, see [1], [3], [4], [5], [6] and references therein. It is highly attractive in that the ILC not only requires little prior knowledge of the control system but also controls the plant effectively.

Traditionally, the applications of the ILC have been focusing on those control systems where a single, repeated operation is performed. This focus includes many industrial systems in manufacturing, robotic system, and chemical engineering areas, where mass production on assembly lines entails repetition. The ILC has been successfully applied to industrial robots [7], industrial batch process control [8], injection-molding machines [9], [10], aluminum extruders [11], trajectory tracking [12], induction motors [13], chain conveyor systems [14], complex stochastic systems [15], nonlinear system [16], rapid thermal processing [17], [18], and semibatch chemical reactors [19].

However, the capacity of resisting disturbance of the ILC needs further improvement [3]. According to [20], the ILC is an integrator along the iterative axis. All the system responses are integrated by the ILC. This means, all the signals including both systems׳ useful knowledge and disturbing signals are “learned” by the ILC. Thus, a disturbing signal will be continuously amplified along the iterative axis as a snowball effect. Up to present, there are many research findings on improving the ILC׳s capacity of disturbance resisting. Most solutions propose a similar kind of ways which adds a filter to the ILC, such as second-order Butterworth filter [20], zero-phase filter [21], or more simply, a forgetting factor. However, for these filter-adding solutions, the filters would filter away certain signals no matter they are useful or disturbing signals.

The main purpose of this paper is to propose a new method to improve the anti-interference ability of the ILC, which is called the IRRM-ILC. The original idea of this new method is to design an intelligent filter which can filter away disturbing signals while not losing system knowledge. The method is achieved by adopting immunological recognition, response, and memory mechanisms, which can separate the useful signals from the disturbing ones while filtering

When it comes to intelligent system with autonomous recognition and differential reaction, the immune system is one of the most typical and effective systems. The immune system can recognize and distinguish the self and non-self organisms of the body, and then eliminate the non-self part successfully. Furthermore, the immune system is also a learning-type system via generating memory cells. Thus, an intelligent algorithm which imitates the recognition, response, and memory mechanisms of immune system will be a feasible method to improve the anti-interference ability of the ILC. Liu et al. [22] proposed a novel reinforcement learning intelligent controller (RLIC) based on immune system mechanism. The method imitated the primary-secondary response of the immune system, which could be regarded as a process of reinforcement learning with memory. It demonstrated that system response ability and stability of the RLIC are better than those of the conventional PID controller.

The main contributions of this paper are as follows. The proposed IRRM-ILC shares the same immunological mechanism of immune system with the RLIC but with different algorithms. The IRRM-ILC is a complementary algorithm that specifically for the ILC considering its data-driven feature. The proposed IRRM-ILC is applied to a temperature control system of wet spinning coagulation processes with a better performance than the conventional ILC.

The paper is organized as follows. In Section 2, the principle and mathematical model of the ILC and immunological mechanism are introduced in details. In Section 3, the structure and steps of the IRRM-ILC method is presented, together with the algorithms of each component of the IRRM-ILC. In Section 4, an application example is given with its simulation results and analysis. Section 5 concludes the whole paper.

Section snippets

The classical iterative learning control algorithm

In this section, the classical ILC algorithm under the continuous time is briefly summarized. The algorithm block diagram is shown in Fig. 1.

The ILC aims at perfectly tracking the desired output of the system without using any information of the state space. Consider the following linear continuous-time system:ẋk(t)=f(xk(t),uk(t),t),yk(t)=g(xk(t),uk(t),t),where xRn×1,yRm×1,uRr×1. f, g are the continuous functions with respect to the state variable x, control input u, and time variable t,

Definition of antigen and antibody

To introduce the IRRM-ILC control algorithm, first, the mathematical expressions of the antigen and antibody are defined as follows.

Application to wet spinning coagulation process

In this section, we use the temperature control system of wet spinning coagulation process as a simulation example to prove the proposed approach is effective.

Conclusions

In this paper, an intelligent iterative learning control method is proposed based on the recognition, response and memory mechanism of immune system. And then, the method is applied to the temperature control system of wet spinning coagulation process. To be specific, the control plant is the vapor heat exchanger in the coagulation bath. A series of simulation experiments are conducted aiming at control the disturbances which are commonly appeared in coagulation bath.

From the simulation results

Nan Xu received his B.S. degree in Electrical Engineering from Donghua University, Shanghai, China, in 2010, where he is currently pursuing his Ph.D. degree in Control Science and Engineering from the College of Information Sciences and Technology. His research interests include intelligent control, chemical process modeling, and multi-objective optimization.

References (31)

  • D.A. Bristow et al.

    A survey of iterative learning control: a learning-based method for high-performance tracking control

    IEEE Control Syst. Mag.

    (2006)
  • H.S. Ahn et al.

    Iterative learning control: survey and categorization

    IEEE Trans. Syst. Man Cybern. C

    (2007)
  • M. Norrlöf

    An adaptive iterative learning control algorithm with experiments on an industrial robot

    IEEE Trans. Robot. Automat.

    (2002)
  • H. Havlicsek et al.

    Nonlinear control of an electrohydraulic injection molding machine via iterative adaptive learning

    IEEE/ASME Trans. Mechatron.

    (1999)
  • M. Pandit et al.

    Optimizing iterative learning control of cyclic production processes with application to extruders

    IEEE Trans. Control Syst. Technol.

    (1999)
  • Cited by (15)

    • An adaptive immune algorithm for service-oriented agricultural Internet of Things

      2019, Neurocomputing
      Citation Excerpt :

      The Artificial Immune System (AIS) is inspired from immunology and acts as an adaptive system that mimics the function of immunology to solve optimization problems [15,16]. For example, Xu et al. [17] used the recognition, response, and memory mechanism of immune system to realize an iterative learning control. Ding et al. [18] applied the working mechanism of the biological immune system to the event-driven dynamic clustering algorithm for wireless sensor networks.

    • Immune intelligent online modeling for the stretching process of fiber

      2019, Information Sciences
      Citation Excerpt :

      The immunological response mechanism is shown in Fig. 1 [23]. Immunological memory stores all the antibodies and ever appeared antigens, and it can do data operating, such as writing, reading and deleting [41]. LPV model is considered as the process model.

    • An endocrine-immune system inspired controllable information diffusion model in social networks

      2018, Neurocomputing
      Citation Excerpt :

      As a part of results, many bio-inspired communication protocols and algorithms have been proposed [21]. In the last decade, the artificial immune system (AIS) has drawn significant attention [22–25] and been successfully applied into many fields [26–28], such as fault detection and diagnosis, web service composition and management, and industrial control. The type and quantity of antigens are far more than the capacity of the antibody's memory, but the immune mechanism can help the antibodies to identify invading of antigens.

    • An immune system-inspired rescheduling algorithm for workflow in Cloud systems

      2016, Knowledge-Based Systems
      Citation Excerpt :

      Throughout the lifetime of an individual, these memory cells remember each specific pathogen encountered and will mount a strong response if the pathogen is detected again. In last decades, artificial immune system has drawn rapidly development and widely applied in many domains, which is out scope of this paper and can be referred to [21-34]. The major functions of the immune system can be summarized as follows:

    • Data-driven controller design for general MIMO nonlinear systems via virtual reference feedback tuning and neural networks

      2016, Neurocomputing
      Citation Excerpt :

      The idea of iterations also has good applications in data-driven method. For instance, iterative learning control (ILC) [9–14] suits for the systems when the off-line data can be obtained repeatedly or periodically and iterative feedback tuning (IFT) developed by Hjalmarsson [15–17] is based on an iterative gradient descent approach. Additionally, in the field of optimal control, adaptive dynamic programming (ADP) [18–21] is a significant and hot topic.

    View all citing articles on Scopus

    Nan Xu received his B.S. degree in Electrical Engineering from Donghua University, Shanghai, China, in 2010, where he is currently pursuing his Ph.D. degree in Control Science and Engineering from the College of Information Sciences and Technology. His research interests include intelligent control, chemical process modeling, and multi-objective optimization.

    Dr. Yongsheng Ding (M׳00–SM′05) is currently a Professor at College of Information Sciences and Technology, Donghua University, Shanghai, China. He obtained his B.S., M.S. and Ph.D. degrees in Electrical Engineering from Donghua University, Shanghai, China in 1989, 1994 and 1998, respectively. From 1996 to 1998, he was a visiting scientist at the Biomedical Engineering Center, The University of Texas Medical Branch, TX, USA. From February 2005 to April 2005, he was a Visiting Professor at Department of Electrical and Computer Engineering, Wayne State University, MI, USA. From September 2007 to February 2008, he was a Visiting Professor at Harvard Medical School, Harvard University, MA, USA. He serves as Senior Member of Institute of Electrical and Electronics Engineers (IEEE). He has published more than 300 technical papers, and five research monograph/advanced textbooks. His scientific interests include computational intelligence, network intelligence, nature-inspired technologies, intelligent robots, internet of things, bio-informatics, and digitized textile technology.

    Dr. Kuangrong Hao is currently a Professor at the College of Information Sciences and Technology, Donghua University, Shanghai, China. She obtained her B.S. degree in Mechanical Engineering from the Hebei University of Technology, Tianjin, China in 1984, her M.S. degree from Ecole Normale Supérieur de Cachan, Paris, France in 1991, and her Ph.D.degree in Mathematics and Computer Science from Ecole Nationale des Ponts et Chaussées, Paris, France in 1995. She has published more than 100 technical papers, and two research monographs. Her scientific interests include machine vision and image processing, robot control, intelligent control, digitized textile technology.

    This work was supported in part by the Key Project of the National Natural Science Foundation of China (No. 61134009), Program for Changjiang Scholars and Innovation Research Team in University from the Ministry of Education (No. IRT1220), Specialized Research Fund for Shanghai Leading Talents, Project of the Shanghai Committee of Science and Technology (Nos. 13JC1407500, and 11JC1400200), Innovation Program of Shanghai Municipal Education Commission (No. 14ZZ067), and the Fundamental Research Funds for the Central Universities (No. 2232012A3-04).

    View full text