Mean-field optimal control

DOI10.1051/COCV/2014009MaRDI QIDQ2926570zbMATH OpenOpenAlexFDO

Authors Massimo Fornasier, Francesco Solombrino

Publication date 31 October 2014

Published in European Series in Applied and Industrial Mathematics (ESAIM): Control, Optimization and Calculus of Variations (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1306.5913

zbMATH Keywords

\(\Gamma\)-limit cost functional mean-field limit ODE constraints PDE constraints sparse optimal control

Mathematics Subject Classification ID

Vlasov equations (35Q83) PDEs in connection with game theory, economics, social and behavioral sciences (35Q91) Dynamical systems in control (37N35) Existence theories for optimal control problems involving ordinary differential equations (49J15) Existence theories for optimal control problems involving partial differential equations (49J20) Methods involving semicontinuity and convergence; relaxation (49J45)

Abstract: We introduce the concept of {it mean-field optimal control} which is the rigorous limit process connecting finite dimensional optimal control problems with ODE constraints modeling multi-agent interactions to an infinite dimensional optimal control problem with a constraint given by a PDE of Vlasov-type, governing the dynamics of the probability distribution of interacting agents. While in the classical mean-field theory one studies the behavior of a large number of small individuals {it freely interacting} with each other, by simplifying the effect of all the other individuals on any given individual by a single averaged effect, we address the situation where the individuals are actually influenced also by an external {it policy maker}, and we propagate its effect for the number

N

of individuals going to infinity. On the one hand, from a modeling point of view, we take into account also that the policy maker is constrained to act according to optimal strategies promoting its most parsimonious interaction with the group of individuals. This will be realized by considering cost functionals including

L^{1}

-norm terms penalizing a broadly distributed control of the group, while promoting its sparsity. On the other hand, from the analysis point of view, and for the sake of generality, we consider broader classes of convex control penalizations. In order to develop this new concept of limit rigorously, we need to carefully combine the classical concept of mean-field limit, connecting the finite dimensional system of ODE describing the dynamics of each individual of the group to the PDE describing the dynamics of the respective probability distribution, with the well-known concept of

G a m m a

-convergence to show that optimal strategies for the finite dimensional problems converge to optimal strategies of the infinite dimensional problem.

Review text

In this very important paper the authors study the discrete (finite-dimensional-continuum) infinite dimensional limit for \(N \rightarrow \infty\) of ODE constrained control problems of the type:NEWLINENEWLINE(1) \(\dot{x}_i = v_i, \dot{v}_i = (H * \mu_N)(x_i v_i)+ f(t, x_i, v_i), i=1,\dots,N, t \in [0,T]\),NEWLINENEWLINEwhere \(\mu_N=\frac{1}{N} \sum^N_{j=1} S_{(x_i v_i)}\), is the empirical atomic measure supported on the agents states \((x_i, v_i)\in \mathbb R^{2d}\), controlled by the minimizer of the cost functionalNEWLINENEWLINE(2) \(\varepsilon^N_v (f): = \int ^T_0 \int_{\mathbb R^{2d}} (L(x,v,\mu_N (t)) + \psi (f(t,x,v)))d \mu_N (t) (x,v)dt\).NEWLINENEWLINEThe authors prove the existence of controls for (1) and (2) based on compactness arguments for the considered class of feedback control functions. The novelty with respect to the usual closed loop control problems stems precisely from the feedback from the control in terms of a locally Lipschitz continuous function \(f(t,.,.)\) of the state variables \((x_j, v_j)\) for \(j=1,2,\dots,N\) in order to grasp, although only intuitively, this fundamental difference.NEWLINENEWLINEThe main result of the proposed work is to clarify in which sense the finite dimensional solutions of (1) and (2) converge for \(N \rightarrow \infty\) to a solution of the PDE constrained problemNEWLINENEWLINE(3) \((\partial \mu / \partial t) + v \cdot \nabla_x \mu = \nabla_v \cdot [H * \mu + f) \mu]\),NEWLINENEWLINEcontrolled by the minimizer \(f\) of the cost functionalNEWLINENEWLINE(4) \(\varepsilon_\psi (f): = \int ^T_0 \int_{\mathbb R^{2d}} (L(x,v,\mu (t)) + \psi (f(t,x,v)))d \mu (t) (x,v)dt\).NEWLINENEWLINEThe authors' arguments are based on the combination of the concept of mean-field, using techniques of optimal transport in order to connect (1) to (3) and the concept of \(\Gamma\)-limit in order to connect the minimizations of (2) and (4) (mean field optimal control).

Recommendations

Cited in

(only showing first 100 items - show all)

This page was built for publication: Mean-field optimal control

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2926570)