Dissipativity learning control (DLC): Theoretical foundations of input-output data-driven model-free control

Abstract Data-driven, model-free control strategies leverage statistical or learning techniques to design controllers based on data instead of dynamic models. We have previously introduced the dissipativity learning control (DLC) method, where the dissipativity property is learned from the input–output trajectories of a system, based on which L 2 -optimal P/PI/PID controller synthesis is performed. In this work, we analyze the statistical conditions on dissipativity learning that enable control performance guarantees, and establish theoretical results on performance under nominal conditions as well as in the presence of statistical errors. The implementation of DLC is further formalized and is illustrated on a two-phase chemical reactor, along with a comparison to model identification-based LQG control.

[1]  Prodromos Daoutidis,et al.  Distributed adaptive dynamic programming for data-driven optimal control , 2018, Syst. Control. Lett..

[2]  B. Brogliato,et al.  Dissipative Systems Analysis and Control , 2000 .

[3]  Jan Maximilian Montenbruck,et al.  Sampling Strategies for Data-Driven Inference of Input–Output System Properties , 2019, IEEE Transactions on Automatic Control.

[4]  J. Willems.  Dissipative dynamical systems Part II: Linear systems with quadratic supply rates , 1972 .

[5]  P. Moylan,et al.  Stability criteria for large-scale systems , 1978 .

[6]  R. Bhushan Gopaluni,et al.  Deep Reinforcement Learning for Process Control: A Primer for Beginners , 2020, ArXiv.

[7]  Svante Gunnarsson,et al.  Iterative feedback tuning: theory and applications , 1998 .

[8]  J. Willems.  Dissipative dynamical systems part I: General theory , 1972 .

[9]  B. Schutter,et al.  Minimal state-space realization in linear system theory: an overview , 2000 .

[10]  Jie Bao,et al.  Dissipativity analysis for networks of process systems , 2013, Comput. Chem. Eng..

[11]  Sergio M. Savaresi,et al.  Virtual reference feedback tuning: a direct method for the design of feedback controllers , 2002, Autom..

[12]  Prodromos Daoutidis,et al.  Dissipativity learning control (DLC): A framework of input-output data-driven control , 2019, Comput. Chem. Eng..

[13]  Pietro Tesi,et al.  Formulas for Data-Driven Control: Stabilization, Optimality, and Robustness , 2019, IEEE Transactions on Automatic Control.

[14]  P. Daoutidis,et al.  Feedback control of nonlinear differential-algebraic-equation systems , 1995 .

[15]  M. Kanat Camlibel,et al.  Data Informativity: A New Perspective on Data-Driven Analysis and Control , 2019, IEEE Transactions on Automatic Control.

[16]  Lennart Ljung,et al.  System Identification: Theory for the User , 1987 .

[17]  Rui Nian,et al.  A review On reinforcement learning: Introduction and applications in industrial process control , 2020, Comput. Chem. Eng..

[18]  Igor Mezic,et al.  Linear predictors for nonlinear dynamical systems: Koopman operator meets model predictive control , 2016, Autom..

[19]  Murat Arcak,et al.  Networks of Dissipative Systems: Compositional Certification of Stability, Performance, and Safety , 2016 .

[20]  P. Rapisarda,et al.  On Lyapunov functions and data-driven dissipativity , 2017 .

[21]  Gábor Szederkényi,et al.  Local dissipative Hamiltonian description of reversible reaction networks , 2008, Syst. Control. Lett..

[22]  P. Moylan,et al.  The stability of nonlinear dissipative systems , 1976 .

[23]  Prodromos Daoutidis,et al.  Input-output data-driven control through dissipativity learning , 2019, 2019 American Control Conference (ACC).

[24]  Françoise Couenne,et al.  The port Hamiltonian approach to modeling and control of Continuous Stirred Tank Reactors , 2011 .

[25]  Shun-ichi Azuma,et al.  Efficient Data-Driven Estimation of Passivity Properties , 2019, IEEE Control Systems Letters.

[26]  Håkan Hjalmarsson,et al.  From experiment design to closed-loop control , 2005, Autom..

[27]  Frank Allgöwer,et al.  One-Shot Verification of Dissipativity Properties From Input–Output Data , 2019, IEEE Control Systems Letters.

[28]  Antonio A. Alonso,et al.  Process systems and passivity via the Clausius-Planck inequality , 1997 .

[29]  Frank Allgöwer,et al.  Determining dissipation inequalities from input-output samples , 2017 .

[30]  Miroslav Krstic,et al.  Nonlinear and adaptive control de-sign , 1995 .

[31]  Alessandro Astolfi,et al.  Stabilization and Disturbance Attenuation of Nonlinear Systems Using Dissipativity Theory , 2002, Eur. J. Control.

[32]  Michel Gevers,et al.  Identification for Control: From the Early Achievements to the Revival of Experiment Design , 2005, Proceedings of the 44th IEEE Conference on Decision and Control.

[33]  Karl Johan Åström,et al.  PID Controllers: Theory, Design, and Tuning , 1995 .

[34]  Haibo He,et al.  Adaptive Critic Nonlinear Robust Control: A Survey , 2017, IEEE Transactions on Cybernetics.

[35]  Bart De Moor,et al.  A note on persistency of excitation , 2005, 2004 43rd IEEE Conference on Decision and Control (CDC) (IEEE Cat. No.04CH37601).

[36]  Zhuo Wang,et al.  From model-based control to data-driven control: Survey, classification and perspective , 2013, Inf. Sci..

[37]  Bo Wahlberg,et al.  Non-parametric methods for L2-gain estimation using iterative experiments , 2010, Autom..

[38]  YangQuan Chen,et al.  Iterative Learning Control: Convergence, Robustness and Applications , 1999 .

[39]  Frank Allgöwer,et al.  Some problems arising in controller design from big data via input-output methods , 2016, 2016 IEEE 55th Conference on Decision and Control (CDC).

[40]  David J. Hill,et al.  Passivity and passification of nonlinear systems , 2000 .