scispace - formally typeset
Search or ask a question
Journal ArticleDOI

From model-based control to data-driven control: Survey, classification and perspective

01 Jun 2013-Information Sciences (Elsevier)-Vol. 235, Iss: 235, pp 3-35
TL;DR: This paper is a brief survey on the existing problems and challenges inherent in model-based control (MBC) theory, and some important issues in the analysis and design of data-driven control (DDC) methods are here reviewed and addressed.
About: This article is published in Information Sciences.The article was published on 2013-06-01. It has received 828 citations till now.
Citations
More filters
Journal ArticleDOI
TL;DR: Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis.
Abstract: Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

13,246 citations

Journal ArticleDOI
TL;DR: The main objective of this paper is to review and summarize the recent achievements in data-based techniques, especially for complicated industrial applications, thus providing a referee for further study on the related topics both from academic and practical points of view.
Abstract: This paper provides an overview of the recent developments in data-based techniques focused on modern industrial applications. As one of the hottest research topics for complicated processes, the data-based techniques have been rapidly developed over the past two decades and widely used in numerous industrial sectors nowadays. The core of data-based techniques is to take full advantage of the huge amounts of available process data, aiming to acquire the useful information within. Compared with the well-developed model-based approaches, data-based techniques provide efficient alternative solutions for different industrial issues under various operating conditions. The main objective of this paper is to review and summarize the recent achievements in data-based techniques, especially for complicated industrial applications, thus providing a referee for further study on the related topics both from academic and practical points of view. This paper begins with a brief evolutionary overview of data-based techniques in the last two decades. Then, the methodologies only based on process measurements and the model-data integrated techniques will be further introduced. The recent developments for modern industrial applications are, respectively, presented mainly from perspectives of monitoring and control. The new trends of data-based technique as well as potential application fields are finally discussed.

856 citations


Cites background from "From model-based control to data-dr..."

  • ...The system can be expressed by three representative dynamic linearization data models, with uniformly bounded PPD [17], [19]....

    [...]

  • ...Model-free adaptive control (MFAC) was proposed by Hou [18], in which a dynamic linearization data model with pseudopartial derivative (PPD) is employed to replace the general nonlinear discrete time system....

    [...]

  • ...More importantly, PPD can be estimated online directly based on the input–output measurements....

    [...]

  • ...Full form of dynamic linearization: Δy(k + 1) = Ψ̃(k)Δũ(k) (20) where Ψ̃(k)= [ Ψ1(k) · · · ΨLu(k)ΨLu+1(k) · · · ΨLy+Lu(k) ] (21) Δũ(k)= [ ΔuT (k), . . . , ΔuT (k − Lu + 1),ΔyT (k), . . . , ΔyT (k − Ly + 1) ]T (22) and its PPD matrix is similar to that of the previous data model....

    [...]

  • ...Partial form of dynamic linearization: Δy(k + 1) = Ψ̄(k)Δū(k) (17) where Ψ̄(k) = [Ψ1(k) · · · ΨL(k) ] , ∥∥Ψ̄(k)∥∥ b (18) Δū(k) = [ΔuT (k), . . . , ΔuT (k − L+ 1) ]T (19) and each Ψi(k) (i = 1, . . . , L) has the structure in (16) to form the PPD matrix....

    [...]

Proceedings ArticleDOI
25 Jun 2019
TL;DR: In this paper, a data-enabled predictive control (DeePC) algorithm is presented that computes optimal and safe control policies using real-time feedback driving the unknown system along a desired trajectory while satisfying system constraints.
Abstract: We consider the problem of optimal trajectory tracking for unknown systems. A novel data-enabled predictive control (DeePC) algorithm is presented that computes optimal and safe control policies using real-time feedback driving the unknown system along a desired trajectory while satisfying system constraints. Using a finite number of data samples from the unknown system, our proposed algorithm uses a behavioural systems theory approach to learn a non-parametric system model used to predict future trajectories. The DeePC algorithm is shown to be equivalent to the classical and widely adopted Model Predictive Control (MPC) algorithm in the case of deterministic linear time-invariant systems. In the case of nonlinear stochastic systems, we propose regularizations to the DeePC algorithm. Simulations are provided to illustrate performance and compare the algorithm with other methods.

411 citations

01 Jan 2015
TL;DR: Finds ways in business and industry is working to manage data and reports on applications that support these initiatives.
Abstract: earelivinginaneraofdatadelugeandasaresult,theterm‘‘big data’’ is appearing in many contexts, from meteorol-ogy, genomics, complex physics simulations, biologicaland environmental research, finance and business tohealthcare. One interesting example is that a press release of SAP AG, dated 11June 2014 reported, ‘‘SAP and the German Football Association turn big dataintosmartdecisionstoimproveplayerperformanceattheWorldCupinBrazil.’’An International Data Corporation (IDC) report [1] predicts that ‘‘from 2005to 2020, the digital universe will grow by a factor of 300, from 130 Exabyte to40 000 Exabyte’’ and that ‘‘from now until 2020 will about double every twoyears.’’ As the name implies, big data literally means large collections of datasets containing abundant information. However, it has some special charac-teristics that distinguish it from ‘‘very large data’’ or ‘‘massive data’’ that aresimply enormous collections of simple-format records, typically equivalent toenormous spreadsheets. Big data, being generally unstructured and heteroge-neous, is extremely complex to dealwith via traditional approaches, andrequires real-time or almost real-timeanalysis. A short definition can there-fore be that ‘‘big data’’ refers to datasets whose size is beyond the abilityof typical database software tools tocapture, store, manage, and analyze.For a thorough discussion on variousaspects of big data and the challengesit presents, together with some po-tential research directions, the readeris referred to [2].

397 citations


Cites background from "From model-based control to data-dr..."

  • ...Methods and algorithms focusing on data-based aspects like statistical analyses in business, management, and biomedicine [10–12], datadriven process monitoring/ prognostics [13], [14], and system control and optimization [15], [16] have been widely investigated in recent...

    [...]

  • ...[16] Z....

    [...]

Journal ArticleDOI
TL;DR: The presented results provide the first (theoretical) analysis of closed-loop properties, resulting from a simple, purely data-driven MPC scheme, including a slack variable with regularization in the cost.
Abstract: We propose a robust data-driven model predictive control (MPC) scheme to control linear time-invariant systems. The scheme uses an implicit model description based on behavioral systems theory and past measured trajectories. In particular, it does not require any prior identification step, but only an initially measured input–output trajectory as well as an upper bound on the order of the unknown system. First, we prove exponential stability of a nominal data-driven MPC scheme with terminal equality constraints in the case of no measurement noise. For bounded additive output measurement noise, we propose a robust modification of the scheme, including a slack variable with regularization in the cost. We prove that the application of this robust MPC scheme in a multistep fashion leads to practical exponential stability of the closed loop w.r.t. the noise level. The presented results provide the first (theoretical) analysis of closed-loop properties, resulting from a simple, purely data-driven MPC scheme.

381 citations


Cites background from "From model-based control to data-dr..."

  • ..., stability or constraint satisfaction of system variables [1], [2]....

    [...]

References
More filters
Book
01 Jan 1987
TL;DR: Das Buch behandelt die Systemidentifizierung in dem theoretischen Bereich, der direkte Auswirkungen auf Verstaendnis and praktische Anwendung der verschiedenen Verfahren zur IdentifIZierung hat.
Abstract: Das Buch behandelt die Systemidentifizierung in dem theoretischen Bereich, der direkte Auswirkungen auf Verstaendnis und praktische Anwendung der verschiedenen Verfahren zur Identifizierung hat. Da ...

20,436 citations


"From model-based control to data-dr..." refers methods in this paper

  • ...However, the model uncertainty descriptions upon which robust control design methods have been based are not consistent with the methods delivered by physical mathematical modeling and identification modeling [86]....

    [...]

Book ChapterDOI
01 Jan 2001
TL;DR: In this paper, the clssical filleting and prediclion problem is re-examined using the Bode-Shannon representation of random processes and the?stat-tran-sition? method of analysis of dynamic systems.
Abstract: The clssical filleting and prediclion problem is re-examined using the Bode-Shannon representation of random processes and the ?stat-tran-sition? method of analysis of dynamic systems. New result are: (1) The formulation and Methods of solution of the problm apply, without modification to stationary and nonstationary stalistics end to growing-memory and infinile -memory filters. (2) A nonlinear difference (or differential) equalion is dericed for the covariance matrix of the optimal estimalion error. From the solution of this equation the coefficients of the difference, (or differential) equation of the optimal linear filter are obtained without further caleulations. (3) Tke fillering problem is shoum to be the dual of the nois-free regulator problem. The new method developed here, is applied to do well-known problems, confirming and extending, earlier results. The discussion is largely, self-contatained, and proceeds from first principles; basic concepts of the theory of random processes are reviewed in the Appendix.

15,391 citations


"From model-based control to data-dr..." refers background in this paper

  • ...Its main branches, system identification, adaptive control, robust control, optimal control, variable structure control, and stochastic system theory, have been extensively used in industrial processes, aerospace, traffic systems, and other applications....

    [...]

Journal ArticleDOI
TL;DR: Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis.
Abstract: Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

13,246 citations

Journal ArticleDOI
TL;DR: This paper presents and proves in detail a convergence theorem forQ-learning based on that outlined in Watkins (1989), showing that Q-learning converges to the optimum action-values with probability 1 so long as all actions are repeatedly sampled in all states and the action- values are represented discretely.
Abstract: \cal Q-learning (Watkins, 1989) is a simple way for agents to learn how to act optimally in controlled Markovian domains. It amounts to an incremental method for dynamic programming which imposes limited computational demands. It works by successively improving its evaluations of the quality of particular actions at particular states. This paper presents and proves in detail a convergence theorem for \cal Q-learning based on that outlined in Watkins (1989). We show that \cal Q-learning converges to the optimum action-values with probability 1 so long as all actions are repeatedly sampled in all states and the action-values are represented discretely. We also sketch extensions to the cases of non-discounted, but absorbing, Markov environments, and where many \cal Q values can be changed each iteration, rather than just one.

8,450 citations