Turnpike in Lipschitz-nonlinear optimal control

We present a new proof of the turnpike property for nonlinear optimal control problems, when the running target is a steady control-state pair of the underlying dynamics. Our strategy combines the construction of suboptimal quasi-turnpike trajectories via controllability, and a bootstrap argument, and does not rely on analyzing the optimality system or linearization techniques. This in turn allows us to address several optimal control problems for finite-dimensional, control-affine systems with globally Lipschitz (possibly nonsmooth) nonlinearities, without any smallness conditions on the initial data or the running target. These results are motivated by the large-layer regime of residual neural networks, commonly used in deep learning applications. We show that our methodology is applicable to controlled PDEs as well, such as the semilinear wave and heat equation with a globally Lipschitz nonlinearity, once again without any smallness assumptions.

[1]  Enrique Zuazua,et al.  Control under constraints for multi-dimensional reaction-diffusion monostable and bistable equations , 2019, 1912.13066.

[2]  K. Kunisch,et al.  Optimal control of an energy-critical semilinear wave equation in 3D with spatially integrated control constraints , 2019, Journal de Mathématiques Pures et Appliquées.

[3]  T. Cazenave.  An introduction to semilinear elliptic equations , 2018 .

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

[6]  Enrique Zuazua,et al.  Long Time versus Steady State Optimal Control , 2013, SIAM J. Control. Optim..

[7]  Enrique Zuazua,et al.  Remarks on Long Time Versus Steady State Optimal Control , 2016 .

[8]  Martin Gugat,et al.  On the Turnpike Phenomenon for Optimal Boundary Control Problems with Hyperbolic Systems , 2018, SIAM J. Control. Optim..

[9]  Karine Beauchard,et al.  Unexpected quadratic behaviors for the small-time local null controllability of scalar-input parabolic equations , 2017, Journal de Mathématiques Pures et Appliquées.

[10]  K. L. Balc'h.  Local controllability of reaction-diffusion systems around nonnegative stationary states , 2018, ESAIM: Control, Optimisation and Calculus of Variations.

[11]  G. Weiss,et al.  Observation and Control for Operator Semigroups , 2009 .

[12]  Stefanie Jegelka,et al.  ResNet with one-neuron hidden layers is a Universal Approximator , 2018, NeurIPS.

[13]  George Cybenko,et al.  Approximation by superpositions of a sigmoidal function , 1989, Math. Control. Signals Syst..

[14]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[15]  Borjan Geshkovski.  Null-controllability of perturbed porous medium gas flow , 2020, ESAIM: Control, Optimisation and Calculus of Variations.

[16]  Enrique Zuazua,et al.  Exact controllability for semilinear wave equations in one space dimension , 1993 .

[17]  P. Cannarsa,et al.  Multiplicative controllability for semilinear reaction-diffusion equations with finitely many changes of sign , 2015, 1510.04203.

[18]  Can Zhang,et al.  Steady-State and Periodic Exponential Turnpike Property for Optimal Control Problems in Hilbert Spaces , 2016, SIAM J. Control. Optim..

[19]  Enrique Zuazua,et al.  On the optimality of the observability inequalities for parabolic and hyperbolic systems with potentials , 2008 .

[20]  Xu Zhang,et al.  Exact Controllability for Multidimensional Semilinear Hyperbolic Equations , 2007, SIAM J. Control. Optim..

[21]  Lars Grüne,et al.  Exponential sensitivity and turnpike analysis for linear quadratic optimal control of general evolution equations , 2020 .

[22]  Emmanuel Tr'elat Ljll,et al.  Linear turnpike theorem , 2020, 2010.13605.

[23]  Domenec Ruiz-Balet,et al.  A Fragmentation Phenomenon for a Nonenergetic Optimal Control Problem: Optimization of the Total Population Size in Logistic Diffusive Models , 2021, SIAM J. Appl. Math..

[24]  P. Lions.  On the Existence of Positive Solutions of Semilinear Elliptic Equations , 1982 .

[25]  Romain Joly,et al.  A Note on the Semiglobal Controllability of the Semilinear Wave Equation , 2014, SIAM J. Control. Optim..

[26]  Herbert Amann,et al.  Linear and Quasilinear Parabolic Problems , 2019, Monographs in Mathematics.

[27]  Enrique Zuazua,et al.  Controllability and Observability of Partial Differential Equations: Some Results and Open Problems , 2007 .

[29]  Enrique Zuazua,et al.  Controllability under positivity constraints of semilinear heat equations , 2017, Mathematical Control & Related Fields.

[30]  Enrique Zuazua,et al.  Problem 5.5 Exact controllability of the semi-linear wave equation , 2009 .

[31]  E Weinan,et al.  A Proposal on Machine Learning via Dynamical Systems , 2017, Communications in Mathematics and Statistics.

[32]  Karine Beauchard,et al.  Quadratic obstructions to small-time local controllability for scalar-input systems , 2018 .

[33]  D. L. Russell.  Review: J.-L. Lions, Controlabilité Exacte, Perturbations et Stabilisation de Systèmes Distribués , 1990 .

[34]  Carlos Esteve,et al.  Large-time asymptotics in deep learning , 2020, ArXiv.

[35]  Enrique Zuazua,et al.  The turnpike property in finite-dimensional nonlinear optimal control , 2014, 1402.3263.

[36]  S. Zamorano.  Turnpike Property for Two-Dimensional Navier–Stokes Equations , 2016, Journal of Mathematical Fluid Mechanics.

[37]  Enrique Zuazua,et al.  Large time control and turnpike properties for wave equations , 2017, Annu. Rev. Control..

[38]  Enrique Zuazua,et al.  Optimal Neumann control for the 1D wave equation: Finite horizon, infinite horizon, boundary tracking terms and the turnpike property , 2015, Syst. Control. Lett..