Learning nonlinear state-space models using autoencoders

We propose a methodology for the identification of nonlinear state–space models from input/output data using machine-learning techniques based on autoencoders and neural networks. Our framework simultaneously identifies the nonlinear output and state-update maps of the model. After formulating the approach and providing guidelines for tuning the related hyper-parameters (including the model order), we show its capability in fitting nonlinear models on different nonlinear system identification benchmarks. Performance is assessed in terms of open-loop prediction on test data and of controlling the system via nonlinear model predictive control (MPC) based on the identified nonlinear state–space model. © 2021 Elsevier Ltd. All rights reserved.

[1]  Alberto Bemporad,et al.  From linear to nonlinear MPC: bridging the gap via the real-time iteration , 2020, Int. J. Control.

[2]  Ole Winther,et al.  A Disentangled Recognition and Nonlinear Dynamics Model for Unsupervised Learning , 2017, NIPS.

[3]  Danilo Comminiello,et al.  Group sparse regularization for deep neural networks , 2016, Neurocomputing.

[4]  Travis E. Oliphant,et al.  Python for Scientific Computing , 2007, Computing in Science & Engineering.

[5]  D. Piga,et al.  Model Structures and Fitting Criteria for System Identification with Neural Networks , 2019, 2020 IEEE 14th International Conference on Application of Information and Communication Technologies (AICT).

[6]  Steven L. Brunton,et al.  Deep learning for universal linear embeddings of nonlinear dynamics , 2017, Nature Communications.

[7]  Richard Hans Robert Hahnloser,et al.  Digital selection and analogue amplification coexist in a cortex-inspired silicon circuit , 2000, Nature.

[8]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[9]  Xinglong Zhang,et al.  A deep learning framework based on Koopman operator for data-driven modeling of vehicle dynamics , 2020, ArXiv.

[10]  Frank Noé,et al.  Time-lagged autoencoders: Deep learning of slow collective variables for molecular kinetics , 2017, The Journal of chemical physics.

[11]  Geoffrey E. Hinton,et al.  Reducing the Dimensionality of Data with Neural Networks , 2006, Science.

[12]  Clarence W. Rowley,et al.  Linearly-Recurrent Autoencoder Networks for Learning Dynamics , 2017, SIAM J. Appl. Dyn. Syst..

[13]  Biao Huang,et al.  Identification of Hammerstein systems without explicit parameterisation of non-linearity , 2009, Int. J. Control.

[14]  Song Han,et al.  Deep Compression: Compressing Deep Neural Network with Pruning, Trained Quantization and Huffman Coding , 2015, ICLR.

[15]  Marco Forgione,et al.  Continuous-time system identification with neural networks: model structures and fitting criteria , 2021, Eur. J. Control.

[16]  Lennart Ljung,et al.  Kernel methods in system identification, machine learning and function estimation: A survey , 2014, Autom..

[17]  Vikas Chandan,et al.  Physics-constrained Deep Learning of Multi-zone Building Thermal Dynamics , 2020, Energy and Buildings.

[18]  Thomas B. Schön,et al.  Deep Learning and System Identification , 2020, IFAC-PapersOnLine.

[19]  Jean-Philippe Noël,et al.  Three Benchmarks Addressing Open Challenges in Nonlinear System Identification , 2017 .

[20]  Alberto Bemporad,et al.  A Bounded-Variable Least-Squares Solver Based on Stable QR Updates , 2020, IEEE Transactions on Automatic Control.

[21]  Victor M. Zavala,et al.  Image-Based Model Predictive Control via Dynamic Mode Decomposition , 2020, Journal of Process Control.

[22]  Andrew R. Barron,et al.  Universal approximation bounds for superpositions of a sigmoidal function , 1993, IEEE Trans. Inf. Theory.

[23]  Eleni Chatzi,et al.  On the use of Nonlinear Normal Modes for Nonlinear Reduced Order Modelling , 2020, XI International Conference on Structural Dynamics.

[24]  Isabelle Guyon,et al.  An Introduction to Feature Extraction , 2006, Feature Extraction.

[25]  G. V. Puskorius,et al.  Truncated backpropagation through time and Kalman filter training for neurocontrol , 1994, Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94).

[26]  Yu Wang,et al.  A new concept using LSTM Neural Networks for dynamic system identification , 2017, 2017 American Control Conference (ACC).

[27]  Maarten Schoukens,et al.  Nonlinear state-space identification using deep encoder networks , 2020, L4DC.

[28]  Alberto Bemporad,et al.  Piecewise affine regression via recursive multiple least squares and multicategory discrimination , 2016, Autom..

[29]  Tao Yang,et al.  Deep Koopman model predictive control for enhancing transient stability in power grids , 2020, International Journal of Robust and Nonlinear Control.

[30]  Ioannis G Kevrekidis,et al.  Extended dynamic mode decomposition with dictionary learning: A data-driven adaptive spectral decomposition of the Koopman operator. , 2017, Chaos.

[31]  Diederik P. Kingma,et al.  An Introduction to Variational Autoencoders , 2019, Found. Trends Mach. Learn..

[32]  Kurt Hornik,et al.  Neural networks and principal component analysis: Learning from examples without local minima , 1989, Neural Networks.

[33]  Jörg Stückler,et al.  Learning to Identify Physical Parameters from Video Using Differentiable Physics , 2020, GCPR.

[34]  Lennart Ljung,et al.  Nonlinear System Identification: A User-Oriented Road Map , 2019, IEEE Control Systems.

[35]  David Q. Mayne,et al.  Model predictive control: Recent developments and future promise , 2014, Autom..

[36]  Steven L. Brunton,et al.  Data-driven discovery of coordinates and governing equations , 2019, Proceedings of the National Academy of Sciences.

[37]  Yoshua Bengio,et al.  A Recurrent Latent Variable Model for Sequential Data , 2015, NIPS.

[38]  Yoram Baram,et al.  Minimal order representation, estimation and feedback of continuous-time stochastic linear systems , 1984 .

[39]  Paul J. Werbos,et al.  Backpropagation Through Time: What It Does and How to Do It , 1990, Proc. IEEE.

[40]  Tadahiro Taniguchi,et al.  Dreaming: Model-based Reinforcement Learning by Latent Imagination without Reconstruction , 2020, ArXiv.

[41]  Uri Shalit,et al.  Deep Kalman Filters , 2015, ArXiv.

[42]  Jasper Snoek,et al.  Practical Bayesian Optimization of Machine Learning Algorithms , 2012, NIPS.

[43]  Maximilian Karl,et al.  Deep Variational Bayes Filters: Unsupervised Learning of State Space Models from Raw Data , 2016, ICLR.

[44]  Alberto Bemporad,et al.  Learning Nonlinear State-Space Models Using Deep Autoencoders , 2018, 2018 IEEE Conference on Decision and Control (CDC).

[45]  Maarten Schoukens,et al.  Non-linear State-space Model Identification from Video Data using Deep Encoders , 2021, IFAC-PapersOnLine.

[46]  Jan Peters,et al.  Stable reinforcement learning with autoencoders for tactile and visual data , 2016, 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

[47]  Matthias W. Seeger,et al.  Deep State Space Models for Time Series Forecasting , 2018, NeurIPS.

[48]  Martin A. Riedmiller,et al.  Embed to Control: A Locally Linear Latent Dynamics Model for Control from Raw Images , 2015, NIPS.

[49]  MORITZ DIEHL,et al.  A Real-Time Iteration Scheme for Nonlinear Optimization in Optimal Feedback Control , 2005, SIAM J. Control. Optim..

[50]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[51]  Naoya Takeishi,et al.  Learning Koopman Invariant Subspaces for Dynamic Mode Decomposition , 2017, NIPS.

[52]  Alberto Bemporad.  Global optimization via inverse distance weighting and radial basis functions , 2020, Comput. Optim. Appl..

[53]  R. Tibshirani.  Regression Shrinkage and Selection via the Lasso , 1996 .

[54]  Lennart Ljung,et al.  Deep State Space Models for Nonlinear System Identification , 2020, IFAC-PapersOnLine.