mheldman / obstacle Goto Github PK

Library of solvers for the discrete obstacle problem, which can be formulated as the following linear complementarity problem:

Find u in R^N such that

                    -Au >= f
                     u  >=  psi
      (Au + f)(u - psi) = 0,

where A is a discretization of the Laplacian on an N-point grid and f is an N-vector.

The solvers have a particular focus on multigrid methods, which are known to be efficient for solving elliptic PDEs like the Poisson equation. The multigrid-based projected full-approximation scheme method ([1]) and standard monotone multigrid method [2] implementations here are able to quickly solve some example problems with millions or ten of millions of unknowns on a desktop computer.

Most of the library is written in Python using scipy.sparse, while the smoothers for the multigrid methods use extension modules written in C++. The code dealing with the setup, compilation, and installation of the extension modules is mostly borrowed from the library of algebraic multigrid solvers pyamg ([3]).

To install, just run sudo python setup.py install. To test the installation, you can use the test.py file. The test.py file is configured to read input data for the obstacle problem from a .py file (stored in the same folder as test.py) specified at the commandline. For example, the following code would solve the dam problem, specified in obstacle/obstacle/test/dam.py, using an eight-grid V(1,1) cycle of the monotone multigrid method with coarse grid size (3 x 3):

python test.py --solver_type monotone --problem_data dam --coarse_mx 1 --coarse_my 1 --show_residuals true --num_grids 8 --cycle_type V --smoothing_iters 1

[1] Achi Brandt and Colin W. Cryer. Multigrid algorithms for the solution of linear complementarity problems arising from free boundary problems. Siam Journal on Scientific and Statistical Computing, 4(4):655–684, 1983.

[2] Ralf Kornhuber. Monotone multigrid methods for elliptic variational inequalities I. Numerische Mathematik, 1994.

[3] L. N. Olson and J.B. Schroder. PyAMG: Algebraic Multigrid Solvers in Python v4.0. https://github.com/pyamg/pyamg. 2018.