betaFeedback

WorksMathematics

Balancing-based model reduction for output-constrained optimal control of switched linear systems

Abstract

We develop a balanced-truncation (BT) framework for model-order reduction of optimal control problems governed by switched linear systems, with quadratic tracking costs, convex input constraints, and Moreau–Yosida-regularized output constraints. For general Petrov–Galerkin reduced models, we derive an a posteriori bound for the error in the optimal value function, and combine it with BT-based estimates of the dynamical-system reduction error. We show that BT preserves the duality structure underlying the first-order optimality conditions, making first-reduce-then-optimize and first-optimize-then-reduce equivalent. Consequently, reducing the forward system also reduces the associated optimality system, leading to a certified, online-efficient, and online-adaptive approximation of the parameterized solution operator. The reduced model can therefore be reused across different regularization parameters, constraint sets, switching signals, and reference trajectories.

Introduction

We consider the optimal control of switched linear systems (SLS) of the form

Σσ{Eσ(t)x˙(t)=Aσ(t)x(t)+Bσ(t)u(t),x(0)=0,x(t+)=Kσ(t+),σ(t−)x(t−),y(t)=Cσ(t)x(t),(1)\Sigma_{\sigma} \left\{ \begin{aligned} E_{\sigma(t)}\dot{x}(t) &= A_{\sigma(t)}x(t) + B_{\sigma(t)}u(t), \quad x(0)=0,\\ x(t^+) &= K_{\sigma(t^+),\sigma(t^-)}x(t^-),\\ y(t) &= C_{\sigma(t)}x(t), \end{aligned} \right. \tag*{(1)}

where σ:R→K:={1,…,L}\sigma:\mathbb{R}\to\mathcal{K}:=\{1,\ldots,L\} is the external switching signal, which we assume to be an element of the set

P:={σ:R→K∣σ is left-continuous with locally finitely many jumps}.(2)\mathcal{P}:=\{\sigma:\mathbb{R}\to\mathcal{K}\mid\sigma\text{ is left-continuous with locally finitely many jumps}\}. \tag*{(2)}

The symbols x(t)∈Rnx(t)\in\mathbb{R}^{n}, u(t)∈Rmu(t)\in\mathbb{R}^{m}, and y(t)∈Rpy(t)\in\mathbb{R}^{p} denote the state, control (or input), and measured output, respectively. We use x(u)x(u) and y(u)y(u) to emphasize their dependence on the control u∈L2(0,T;Rm)u\in L^{2}(0,T;\mathbb{R}^{m}) with terminal time T>0T>0. For each mode j∈Kj\in\mathcal{K}, the matrices Ej,Aj∈Rn×nE_j,A_j\in\mathbb{R}^{n\times n}, Bj∈Rn×mB_j\in\mathbb{R}^{n\times m}, and Cj∈Rp×nC_j\in\mathbb{R}^{p\times n} define the active subsystem and EjE_j is assumed to be invertible. The matrices Ki,j∈Rn×nK_{i,j}\in\mathbb{R}^{n\times n} describe state transitions from mode jj to mode ii at the switching times, with Ki,i=InK_{i,i}=I_n. As defined in § 1.3, σ(t−)\sigma(t^-) and σ(t+)\sigma(t^+) denote the left and right limits, respectively. We refer to (1.1) as the full-order model (FOM). For each fixed σ∈P\sigma\in\mathcal{P}, let 0=t0<⋯<tN+1=T0 = t_0 < \cdots< t_{N+1} = T be the partition induced by the switching times, i.e., the discontinuities of σ\sigma in the interval (0,T)(0,T). The state and output satisfy x∣(tk,tk+1)∈H1(tk,tk+1;Rn)↪C([tk,tk+1];Rn)\mathbf{x}|_{(t_k,t_{k+1})} \in H^1(t_k,t_{k+1};\mathbb{R}^n) \hookrightarrow C([t_k,t_{k+1}];\mathbb{R}^n) and y∣(tk,tk+1)∈C([tk,tk+1];Rp)y|_{(t_k,t_{k+1})} \in C([t_k,t_{k+1}];\mathbb{R}^p) for k=0,…,Nk = 0, \ldots, N.

Given a cost parameter λ>0\lambda> 0, we consider the finite-horizon optimal control of (1) with respect to the quadratic cost function

J(y,u):=∫0T(12∥y(t)−yd(t)∥Rp2+λ2∥u(t)−ud(t)∥Rm2) dt.(3)J(\mathbf{y},\mathbf{u}) := \int_0^T \left(\frac{1}{2}\lVert y(t)-y_{\mathrm{d}}(t)\rVert_{\mathbb{R}^p}^2 + \frac{\lambda}{2}\lVert u(t)-u_{\mathrm{d}}(t)\rVert_{\mathbb{R}^m}^2\right)\,\mathrm{d}t. \tag*{(3)}

where ud∈L2(0,T;Rm)u_{\mathrm{d}} \in L^2(0,T;\mathbb{R}^m) and yd∈L2(0,T;Rp)y_{\mathrm{d}} \in L^2(0,T;\mathbb{R}^p) are prescribed references. Let Uad⊆RmU_{\mathrm{ad}} \subseteq\mathbb{R}^m and Yad⊆RpY_{\mathrm{ad}} \subseteq\mathbb{R}^p be nonempty, closed, convex sets of admissible inputs and outputs, respectively, and define Uad:=L2(0,T;Uad)\mathcal{U}_{\mathrm{ad}} := L^2(0,T;U_{\mathrm{ad}}). The pointwise output-constrained full-order optimal control problem (FOM-OCP) is

min⁡J(y,u)s.t.(y,u) solves xref‘(1.1)‘,u∈Uad,y(t)∈Yad for all t∈[0,T].(4)\min J(\mathbf{y},\mathbf{u}) \quad\text{s.t.}\quad(\mathbf{y},\mathbf{u})\text{ solves }{xref}`(1.1)`,\quad\mathbf{u}\in\mathcal{U}_{\mathrm{ad}},\quad y(t)\in Y_{\mathrm{ad}}\text{ for all }t\in[0,T]. \tag*{(4)}

By piecewise continuity and closedness of YadY_{\mathrm{ad}}, the output constraints in (4) are equivalent to y(t)∈Yady(t) \in Y_{\mathrm{ad}} for almost every t∈(0,T)t \in(0,T), in particular, both one-sided limits at each interior switching time belong to YadY_{\mathrm{ad}}. A natural setting for multiplier theory is the product of continuous-function spaces on the closed switching intervals, in which Lagrange multipliers consist of tuples of finite vector-valued measures (cf., e.g., [35, Sec. 6.2.1]). To obtain computable multipliers, one can consider a Moreau–Yosida (MY) regularization in L2L^2 (cf. [6], [19, Sec. 2.7.2.1]). To this end, define the nonempty, closed, convex set

Yad:={v∈L2(0,T;Rp)∣v(t)∈Yad for almost every t∈(0,T)}.\mathcal{Y}_{\mathrm{ad}} := \left\{\mathbf{v}\in L^2(0,T;\mathbb{R}^p)\mid v(t)\in Y_{\mathrm{ad}}\text{ for almost every }t\in(0,T)\right\}.

For a regularization parameter γ>0\gamma> 0, the MY approximation is based on the metric projection

PYad:L2(0,T;Rp)→Yad,PYad(y):=arg min⁡v∈Yad∥y−v∥L2(0,T;Rp).P_{\mathcal{Y}_{\mathrm{ad}}}:L^2(0,T;\mathbb{R}^p)\to\mathcal{Y}_{\mathrm{ad}},\qquad P_{\mathcal{Y}_{\mathrm{ad}}}(\mathbf{y}):=\operatorname*{arg\,min}_{\mathbf{v}\in\mathcal{Y}_{\mathrm{ad}}}\lVert\mathbf{y}-\mathbf{v}\rVert_{L^2(0,T;\mathbb{R}^p)}.

With the regularized cost functional Jγ(y,u):=J(y,u)+γ2∥y−PYad(y)∥L2(0,T;Rp)2J_\gamma(\mathbf{y},\mathbf{u}):=J(\mathbf{y},\mathbf{u})+\frac{\gamma}{2}\lVert\mathbf{y}-P_{\mathcal{Y}_{\mathrm{ad}}}(\mathbf{y})\rVert_{L^2(0,T;\mathbb{R}^p)}^2 we obtain the MY-regularized FOM-OCP

min⁡Jγ(y,u)s.t.(y,u) solves xref‘(1.1)‘ and u∈Uad.(5)\min J_\gamma(\mathbf{y},\mathbf{u}) \quad\text{s.t.}\quad(\mathbf{y},\mathbf{u})\text{ solves }{xref}`(1.1)`\text{ and }\mathbf{u}\in\mathcal{U}_{\mathrm{ad}}. \tag*{(5)}

We denote the unique solution of (5) by (y⋆,u⋆)(\mathbf{y}^\star,\mathbf{u}^\star); see Proposition 2.1. The convergence of solutions of (5) towards the solution of (4) as γ→∞\gamma\to\infty is investigated, e.g., in [21]. Approximating the solution of (5) numerically requires repeated evaluations of (1) for different inputs u∈Uad\mathbf{u}\in\mathcal{U}_{\mathrm{ad}}. These evaluations become computationally expensive when nn is large, particularly if the OCP must be solved for many configurations of the parameter tuple

References

  1. [1]A. C. Antoulas, Approximation of Large-Scale Dynamical Systems, Advances in Design and Control, SIAM, Philadelphia, PA, 2005, https://doi.org/10.1137/1.9780898718713.
  2. [2]B. Azmi and M. Berneruther, On the forward-backward method with nonmonotone linesearch for infinite-dimensional nonsmooth nonconvex problems, Comput. Optim. Appl., 91 (2025), pp. 1263–1308, https://doi.org/10.1007/s10589-025-00684-x.
  3. [3]E. Bader, Reduced Basis Methods Applied to Obstacle Problems and Parametrized Distributed Optimal Control Problems with Control and State Constraints, PhD thesis, RWTH Aachen University, Aachen, Germany, 2016, https://publications.rwth-aachen.de/record/674368.DOI
  4. [4]E. Bader, M. Kärcher, M. A. Grepl, and K. Veroy, Certified reduced basis methods for parametrized distributed elliptic optimal control problems with control constraints, SIAM J. Sci. Comput., 38 (2016), pp. A3921–A3946, https://doi.org/10.1137/16M1059898.
  5. [5]M. Baumann, P. Benner, and J. Heiland, Space-time Galerkin POD with application in optimal control of semi-linear parabolic partial differential equations, SIAM J. Sci. Comput., 40 (2018), pp. A1611–A1641, https://doi.org/10.1137/17M1135281.
  6. [6]H. H. Bauschke and P. L. Combettes, Convex Analysis and Monotone Operator Theory in Hilbert Spaces, CMS Books in Mathematics, Springer, New York, 2011, https://doi.org/10.1007/978-1-4419-9467-7.
  7. [7]M. Baştug, M. Petreczky, R. Wisniewski, and J. Leth, Model reduction by nice selections for linear switched systems, IEEE Trans. Automat. Control, 61 (2016), pp. 3422–3437, https://doi.org/10.1109/TAC.2016.2518023.
  8. [8]P. Benner, S. Gugercin, and K. Willcox, A survey of projection-based model reduction methods for parametric dynamical systems, SIAM Rev., 57 (2015), pp. 483–531, https://doi.org/10.1137/130932715.
  9. [9]P. Benner, M. Ohlberger, A. Cohen, and K. Willcox, eds., Model Reduction and Ap- proximation: Theory and Algorithms, vol. 15 of Computational Science & Engineering, Society for Industrial and Applied Mathematics, Philadelphia, PA, 2017, https://doi.org/10.1137/1.9781611974829.
  10. [10]P. Benner, R. Sorensen, and V. Mehrmann, eds., Dimension Reduction of Large-Scale Systems: Proceedings of a Workshop held in Oberwolfach, Germany, October 19-25, 2003, vol. 45 of Lecture Notes in Computational Science and Engineering, Springer, Berlin, Heidelberg, 2005, https://doi.org/10.1007/3-540-27909-1.
  11. [11]T. Breiten and K. Kunisch, Feedback stabilization of the Schlögl model by LQG-balanced truncation, in 2015 European Control Conference (ECC), IEEE, 2015, pp. 1171–1176, https://doi.org/10.1109/ECC.2015.7330698.
  12. [12]S. Dietze and M. A. Grepl, Reduced order model predictive control for parametrized parabolic partial differential equations, Appl. Math. Comput., 453 (2023), p. 128044, https://doi.org/10.1016/j.amc.2023.128044.
  13. [13]D. F. Enns, Model Reduction for Control System Design, PhD thesis, Stanford University, Stanford, California, 1984.

Paper details

vlnc:W_5mjbwdc2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents