# Mixed Linear Complementarity Problems (MLCP)¶

The problem:

Find $$(z,w)$$ such that:

:math: left{ begin{array}{l} M z + q = w \ w_1=0 \ 0 le w_{2} perp v ge 0 end{array} right. text{ with } z= left[ begin{array}{c} u\ v\ end{array} right] text{ and } w= left[ begin{array}{c} w_{1}\ w_{2}\ end{array} right]

:math: u, w_{1} are vectors of size n.

:math: v, w_{2} are vectors of size m.

Another storage is also possible for the MLCP problem:

Try $$(u,v,w)$$ such that:

:math: leftlbrace begin{array}{l} A u + Cv +a =0\ D u + Bv +b = w \ 0 le v perp w ge 0\ end{array} right.

where A is an ( :math: n times n ) matrix, B is an ( :math: m times m ) matrix, C is an ( :math: n times m ) matrix,

D is an ( :math: m times n ) matrix, a and u is an ( :math: n  ) vectors b,v and w is an ( :math: m  ) vectors.

Available solvers:

The solvers and their parameters are described in Mixed Linear Complementarity Problems Solvers.

Use the generic function mlcp_solver(), to call one the the specific solvers listed below:

Note that all the algorithms are not available for the two options of storage M,q or A,B,C,D,a,b

(see the functions/solvers list in MLCP_solvers.h )

# Mixed Linear Complementarity Problems Solvers¶

This page gives an overview of the available solvers for MLCP and their required parameters.

This page gives an overview of the available solvers for MLCP and their required parameters.

For each solver, the input argument are:

PGS Solver:

Projected Gauss-Seidel solver

function: mlcp_pgs()

parameters:

• iparam (in): maximum number of iterations allowed
• iparam (out): number of iterations processed
• iparam (in): 0 for implicit, 1 for explicit
• dparam (in): tolerance
• dparam (out): resulting error

PGS Solver for SBM storage:

Projected Gauss-Seidel solver

• iparam (in): maximum number of iterations allowed for GS process
• iparam (out): number of GS iterations processed
• iparam (out): sum of all local number of iterations (if it has sense for the local solver)
• dparam (in): tolerance
• dparam (out): resulting error
• dparam (in): sum of all local error values

RPGS Solver:

Regularized Projected Gauss-Seidel, solver for MLCP, able to handle with matrices with null diagonal terms

function: mlcp_rpgs()

parameters:

• iparam (in): maximum number of iterations allowed
• iparam (out): number of iterations processed
• dparam (in): tolerance
• dparam (out): resulting error
• dparam (in): rho

PSOR Solver:

Projected Succesive over relaxation solver for MLCP. See cottle, Pang Stone Chap 5 function: mlcp_psor()

parameters:

• iparam (in): maximum number of iterations allowed
• iparam (out): number of iterations processed
• dparam (in): tolerance
• dparam (out): resulting error
• dparam (in): omega

RPSOR Solver:

Regularized Projected Succesive over relaxation solver for MLCP function: mlcp_rpsor()

parameters:

• iparam (in): maximum number of iterations allowed
• iparam (out): number of iterations processed
• dparam (in): tolerance
• dparam (out): resulting error
• dparam (in): omega
• dparam (in): rho

Path (Ferris) Solver:

The path solver must be initialize:

1. Initialize the solver with mlcp_driver_init().
2. Use a lot with mlcp_driver().
3. Reset the solver with mlcp_driver_reset().

function: mlcp_path()

parameters:

• dparam (in): tolerance

ENUM Solver:

The enumeratif solver must be initialize:

1. Initialize the solver with mlcp_driver_init().
2. Use a lot with mlcp_driver().
3. Reset the solver with mlcp_driver_reset().

function: mlcp_enum()

The enumeratif solver must be initialize:

1. Initialize the solver with mlcp_driver_init().
2. Use a lot with mlcp_driver().
3. Reset the solver with mlcp_driver_reset().

parameters:

• dparam (in): a positive value, tolerane about the sign.
• iparam (in) : use DGELS (1) or DGESV (0).
• dWork : working float zone size : The number of doubles is retruned by the function mlcp_driver_get_dwork(). MUST BE ALLOCATED BY THE USER.
• iWork : working int zone size : . The number of double is retruned by the function mlcp_driver_get_iwork(). MUST BE ALLOCATED BY THE USER.

SIMPLEX Solver:

The simplex solver must be initialize:

1. Initialize the solver with mlcp_driver_init().
2. Use a lot with mlcp_driver().
3. Reset the solver with mlcp_driver_reset().

function: mlcp_simplex()

The simplex solver must be initialize:

1. Initialize the solver with mlcp_driver_init().
2. Use a lot with mlcp_driver().
3. Reset the solver with mlcp_driver_reset().

parameters:

• iparam (in): Max number of iteration (example: 1000000).
• dparam (in): A positive value, tolerance to consider that a var is null(ex: 10e-12).
• dparam (in): A positive value, tolerance to consider that complementarity holds(ex: 10e-12).
• dparam (in): A positive value, tolerance to consider that a var is negative(ex: 10e-9).
• dWork : working float zone size : The number of doubles is retruned by the function mlcp_driver_get_dwork(). MUST BE ALLOCATED BY THE USER.
• iWork : working int zone size : . The number of double is retruned by the function mlcp_driver_get_iwork(). MUST BE ALLOCATED BY THE USER.

DIRECT_ENUM Solver:

The direct and enumeratif solver must be initialize:

1. Initialize the solver with mlcp_driver_init().
2. Use a lot with mlcp_driver().
3. Reset the solver with mlcp_driver_reset().

function: mlcp_direct_enum().

parameters:

• iparam (in): Number of registered configurations.
• iparam (out): Number of case the direct solved failed.
• dparam (in): A positive value, tolerane about the sign.
• dparam (in): A tolerance for the direct solver to consider that a var is negative(ex: 1e-12).
• dparam (in): A tolerance for the direct solver to consider that a var is positive(ex: 1e-12).
• dWork : working float zone size : The number of doubles is retruned by the function mlcp_driver_get_dwork(). MUST BE ALLOCATED BY THE USER.
• iWork : working int zone size : . The number of double is retruned by the function mlcp_driver_get_iwork(). MUST BE ALLOCATED BY THE USER.

DIRECT_PATH Solver:

The path solver must be initialize:

1. Initialize the solver with mlcp_driver_init().
2. Use a lot with mlcp_driver().
3. Reset the solver with mlcp_driver_reset().

function: mlcp_direct_path()

parameters:

• iparam (in): Number of registered configurations.
• iparam (out): Number of case the direct solved failed.
• dparam (in): Tolerance.
• dparam (in): A tolerance for the direct solver to consider that a var is negative(ex: 1e-12).
• dparam (in): A tolerance for the direct solver to consider that a var is positive(ex: 1e-12).
• dWork : working float zone size : The number of doubles is retruned by the function mlcp_driver_get_dwork(). MUST BE ALLOCATED BY THE USER.
• iWork : working int zone size : . The number of double is retruned by the function mlcp_driver_get_iwork(). MUST BE ALLOCATED BY THE USER.

DIRECT_SIMPLEX Solver:

The direct and simplex solver must be initialize:

1. Initialize the solver with mlcp_driver_init().
2. Use a lot with mlcp_driver().
3. Reset the solver with mlcp_driver_reset().

function: mlcp_direct_simplex().

parameters:

• iparam (in): Max number of iteration (example: 1000000).
• iparam (in): Number of registered configurations.
• iparam (out): Number of case the direct solved failed.
• dparam (in): A positive value, tolerance to consider that a var is null(ex: 10e-12).
• dparam (in): A positive value, tolerance to consider that complementarity holds(ex: 10e-12).
• dparam (in): A positive value, tolerance to consider that a var is negative(ex: 10e-9).
• dparam (in): A tolerance for the direct solver to consider that a var is negative(ex: 1e-12).
• dparam (in): A tolerance for the direct solver to consider that a var is positive(ex: 1e-12).
• dWork : working float zone size : The number of doubles is retruned by the function mlcp_driver_get_dwork(). MUST BE ALLOCATED BY THE USER.
• iWork : working int zone size : . The number of double is retruned by the function mlcp_driver_get_iwork(). MUST BE ALLOCATED BY THE USER.