Description

This model shows how to solve a simple VI using EMP

VI is to find x in X:
F(x) (y - x) >= 0, for all y in X
X is a closed convex set, e.g. X = { x >= 0 | g(x) >= 0 }

Subsequently, the model 'reform' illustrates the MCP that EMP solves internally.

Contributor: Michael Ferris, February 2010

Small Model of Type : VI

Category : GAMS EMP library

Main file : simplevi.gms

$title Simple Variational Inequality (SIMPLEVI,SEQ=46)

$onText
This model shows how to solve a simple VI using EMP

VI is to find x in X:
F(x) (y - x) >= 0, for all y in X
X is a closed convex set, e.g. X = { x >= 0 | g(x) >= 0 }

Subsequently, the model 'reform' illustrates the MCP that EMP solves internally.

Contributor: Michael Ferris, February 2010
$offText

sets
 I  / i1, i2 /
 J  / j1 * j3 /
 ;

table A(I,J)
       j1        j2     j3
i1      1         1
i2                1      1 ;

parameter b(I) /
i1   6
i2   9
/;

positive variable
 x(J)  'primal vars, perp to f(J)'
 ;

equations
 F(J)
 g(I)
 ;

F(J)..      2 * x(J)             =n= 0    ;
g(I)..      sum {j, A(I,J)*x(J)} =g= b(I) ;

model simpleVI / F, g/;

file fx /"%emp.info%"/;
putclose fx 'vi F x g';

solve simpleVI using emp;


$onText
This could be reformulated as an MCP:

0 <= F(x) - \grad g(x) u  \perp x >= 0
0 <= g(x)                                  \perp u >= 0
$offText

positive variable
 u(I)  'auxiliary vars, perp to g(I)'
 ;

equation
 F_aux(J)
 ;

F_aux(J)..  2 * x(J) + sum{I, -A(I,J)*u(I)} =n= 0 ;
u.l(I) = g.m(I);
model reform / F_aux.x, g.u/;

*benchmark the solution obtained from simpleVI
reform.iterlim = 0;
solve reform using mcp;
abort$(reform.objval > 1e-9) 'Solutions differ'