A Semi-Lagrangian scheme for Hamilton-Jacobi-Bellman equations with Dirichlet boundary conditions

Elisabetta Carlini 1, Athena Picarelli 2, Francisco J. Silva 3 · 2025

Read A Semi-Lagrangian scheme for Hamilton-Jacobi-Bellman equations with Dirichlet boundary conditions by Elisabetta Carlini 1, Athena Picarelli 2, Francisco J. Silva 3 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-20283.txt.

Introduction

Elisabetta Carlini, Athena Picarelli, and Francisco Silva design a numeric scheme for Hamilton-Jacobi-Bellman equations that respects hard bounds. These equations rest on time and can be degenerate. They are of second order and arise in control theory. Their answers are grasped in the viscosity sense. A scheme works if it is monotone, fits the equation, and is stable. Standard finite differences are in most cases not monotone. Semi-Lagrangian schemes are monotone by how they are built.

The catch is that such schemes use a wide stencil. In a bounded domain that stencil tends to step past the edge and break the fit. The team proposes a scheme of this kind on a mesh with no set grid. It treats points near the edge in a special way to keep the fit. They prove the scheme works for problems whose answer can even be discontinuous.

Catalog record

CREATORElisabetta Carlini 1, Athena Picarelli 2, Francisco J. Silva 3
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE102 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-20283

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A Semi-Lagrangian scheme for Hamilton-Jacobi-Bellman equations with Dirichlet boundary conditions Thanks: 1 Dipartimento di Matematica Guido Castelnuovo, “Sapienza”, Università di Roma, 00185 Rome, Italy ([email protected]). Thanks: 2 Dipartimento di Scienze Economiche, Università degli Studi di Verona, 37129 Verona, Italy ([email protected]). Thanks: 3 Institut de recherche XLIM-DMI, UMR-CNRS 7252, Faculté des Sciences et Techniques, Université de Limoges, 87060 Limoges, France ([email protected]). Elisabetta Carlini 1, Athena Picarelli 2, Francisco J. Silva 3 Abstract . We study the numerical approximation of time-dependent, possibly degenerate, second-order Hamilton-Jacobi-Bellman equations in bounded domains with nonhomogeneous Dirichlet boundary conditions. It is well known that convergence towards the exact solution of the equation, considered here in the viscosity sense, holds if the scheme is monotone, consistent, and stable. While standard finite difference schemes are, in general, not monotone, the so-called semi-Lagrangian schemes are monotone by construction. On the other hand, these schemes make use of a wide stencil and, when the equation is set in a bounded domain, this typically causes an overstepping of the boundary and hence the loss of consistency. We propose here a semi-Lagrangian scheme defined on an unstructured mesh, with a suitable treatment at grid points near the boundary to preserve consistency, and show its convergence for problems where the viscosity solution can even be discontinuous. We illustrate the numerical convergence in several tests, including degenerate and first-order equations. AMS-Subject Classification: 49L25, 65M12, 35D40, 35K55. Keywords: Hamilton-Jacobi-Bellman equations, Dirichlet boundary conditions, semi-Lagrangian schemes, convergence results, numerical experiences. 1. Introduction We consider the following second-order, possibly degenerate, Hamilton-Jacobi-Bellman (HJB) equation (1a) \displaystyle-\partial_{t}v+\max_{a\in A}\mathcal{L}^{a}(t,x,D_{x}v,D^{2}_{x}v) \displaystyle=0\qquad\text{in }Q_{T}, (1b) \displaystyle v \displaystyle=\Psi\qquad\text{on }\partial^{*}Q_{T}, where, for a given T>0 and a nonempty open and bounded domain \Omega\subset{\mathbb{R}}^{d} , Q_{T}:=[0,T)\times\Omega and \partial^{*}Q_{T}:=(\{T\}\times\overline{\Omega})\cup([0,T)\times\partial\Omega) denote, respectively, the parabolic domain and its parabolic boundary, \emptyset\neq A\subset{\mathbb{R}}^{m} is compact, \Psi\colon\partial^{*}Q_{T}\to{\mathbb{R}} and, given f\colon\overline{Q}_{T}\times A\to{\mathbb{R}} , b\colon\overline{Q}_{T}\times A\to{\mathbb{R}}^{d} , and \sigma\colon\overline{Q}_{T}\times A\to{\mathbb{R}}^{d\times p} , for every a\in A , (2) \mathcal{L}^{a}(t,x,q,P)=-\frac{1}{2}\text{Tr}[\sigma\sigma^{\top}(t,x,a)P]-b(t,x,a)^{\top}q-f(t,x,a)\quad\text{for all }(t,x,q,P)\in\overline{Q}_{T}\times{\mathbb{R}}^{d}\times{\mathbb{R}}^{d\times d}. Observe that if A=\{a\} for some a\in{\mathbb{R}}^{m} , by considering a simple change of the time variable, equation ( 1 ) reduces to a standard, possibly degenerated, linear parabolic equation. When A is not a singleton, system ( 1 ) is nonlinear and naturally arises in the study of the optimal control of diffusion processes (see, e.g., [ 28 ] ). Under several assumptions on the data, imposing a kind of compatibility condition with the boundary value \Psi , the existence and uniqueness of a continuous viscosity solution to ( 1 ) (see, e.g., [ 20 ] ) can be found, for instance, in [ 24 ] and [ 28 ] for first- and second-order equations, respectively. For more general data, the study of the existence and uniqueness of viscosity solutions satisfying a weak form of the boundary condition has been conducted in [ 5 , 29 , 6 ] for first-order equations, in [ 4 , 7 ] for stationary second-order equations, and in [ 18 ] for a class of time-dependent problems. The numerical approximation of boundary value problems such as ( 1 ) is practically relevant not only when the original problem is posed on a bounded domain, but also when it is defined on the whole space and the localization to a bounded region is required for computational tractability. On the other hand, its numerical analysis is significantly more challenging than the one where \Omega={\mathbb{R}}^{d} , as the regularity of the solution near the boundary \partial\Omega plays an important role. In this context, Godunov-type schemes for first-order stationary and time-dependent HJB equations with nonhomogeneous Dirichlet conditions have been considered in [ 1 ] and [ 45 ] , respectively. A variation of the abstract scheme in [ 1 ] is investigated in [ 45 ] and yields, in particular, the convergence of a finite volume approximation of a time-dependent HJB equation with nonhomogeneous Dirichlet conditions. For second-order stationary HJB equations with homogeneous Dirichlet conditions satisfying the so-called Cordes condition, a convergent discontinuous Galerkin finite element approximation is studied in [ 44 ] . In the case of time-dependent HJB equations with Dirichlet conditions and admitting continuous solutions, [ 28 , Chapter IX] and [ 32 ] show the convergence of monotone finite difference and finite element methods, respectively. Let us also mention the work [ 30 ] dealing with finite element approximation of time-dependent HJB equations with mixed boundary conditions and [ 33 ] , which discusses the relevance of the definition of boundary conditions for HJB equations including several numerical examples. For a more complete overview on existing numerical methods to approximate solutions to fully nonlinear elliptic and parabolic equations with or without boundary conditions, we refer the reader to [ 26 , 40 ] and the references therein. In this work, we study the approximation of ( 1 ) through semi-Lagrangian (SL) schemes, which are traditionally based on the discretization of representation formulae of solutions to PDEs in terms of their characteristic curves. Compared with standard finite difference schemes, one advantage of SL schemes is that they are at the same time explicit and stable under an inverse Courant-Friedrichs-Lewy (CFL) condition and hence allow for large time steps. We refer the reader to [ 25 ] for an overview of SL schemes for first-order HJB equations. When considering equation ( 1 ) with \Omega={\mathbb{R}}^{d} , in [ 38 , 16 , 17 , 21 ] the authors consider a SL discretization that takes advantage of the fact that the viscosity solution is given by the so-called value function of an optimal control problem of a diffusion process evolving in {\mathbb{R}}^{d} . Roughly speaking, they consider a stochastic Euler time-discrete scheme for the underlying controlled characteristic curves combined with interpolation methods to deal with the space variable. Similarly, as recalled in Section 2 below, the solution to ( 1 ) is given by the value function of an optimal control problem of a controlled diffusion where a cost, represented by \Psi(t,x) , is incurred as soon as the controlled trajectory hits the boundary at point x\in\partial\Omega at time t\in[0,T] , and the trajectory is stopped at x until the final time T . If one proceeds as in the case where \Omega={\mathbb{R}}^{d} , a particular difficulty arises at grid points near the boundary \partial\Omega , for which at least one of the possible time-discrete trajectories hits \partial\Omega before the prescribed time step. Indeed, for those points, the SL discretization in [ 38 , 16 , 21 ] looses its consistency. In order to cope with this problem, some extensions of SL schemes for PDEs with Dirichlet boundary conditions have been considered in the literature. In [ 27 , 31 ] , a SL scheme for the Monge-Ampère equation, based on a rescaling of the space mesh near the boundary, is presented. A second-order fully discrete SL scheme, based on an extrapolation technique, is considered in [ 11 ] to solve systems of advection-diffusion-reaction equations. Finally, a SL scheme for second-order HJB equations has been proposed in [ 43 , 41 ] , for which, in order to deal with some instabilities, an implicit version had to be introduced. Inspired by the scheme proposed by Milstein and Tretyakov in [ 39 ] (see also [ 13 ,

Read the complete text