Modified domain decomposition method for Hamilton-Jacobi-Bellman equations

From MaRDI portal
Publication:616035





The authors consider a class of optimal control problems which is connected to a series of second order elliptic equations posed in \(\mathbb R^n\), along with first kind boundary values. After a suitable discretization, they get a discrete Hamilton-Jacobi-Bellman equation and investigate an iterative method for its numerical solution which requires the solution of variational inequalities in every step. Here, a domain decomposition method helps to parallelize the approach. Instead of the Jacobi-type iteration of \textit{S. Zhou} and \textit{W. Zhan} [J. Comput. Appl. Math. 159, No.~1, 195--204 (2003; Zbl 1034.65052)], they consider a Gauss-Seidel type iteration and prove its convergence. They also give numerical results showing the improvement over the earlier approach. Reviewer's remark: It should be noted that there are several difficulties connected mostly to badly chosen or not explained indices hindering a clear understanding of the paper -- the simplest of which consists in taking the same index, \(n\), for the iteration count and for the dimension of the original elliptic problems.











This page was built for publication: Modified domain decomposition method for Hamilton-Jacobi-Bellman equations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q616035)