circlesp.gms : Circle Enclosing Points - Stochastic Example

Description

This is an example from the GAMS/SNOPT manual. Find the smallest circle
that contains a number of given points.


Small Model of Type : SP


Category : GAMS EMP library


Main file : circlesp.gms

$title Circle Enclosing Points - Stochastic Example (CIRCLESP,SEQ=88)
$ontext

This is an example from the GAMS/SNOPT manual. Find the smallest circle
that contains a number of given points.


Gill, P E, Murray, W, and Saunders, M A, GAMS/SNOPT: An SQP Algorithm
for Large-Scale Constrained Optimization, 1988.

$offtext
* DECIS solves only LP models
$ifi %gams.emp%==decis $exit

$if not set size $set size 3
set i points /p1*p%size%/;

parameters
   x(i)    x coordinates,
   y(i)    y coordinates;

* fill with random data
x(i) = uniform(1,10);
y(i) = uniform(1,10);

variables
   a       x coordinate of center of circle
   b       y coordinate of center of circle
   r       radius;

equations
   e(i)    points must be inside circle;


e(i)..  sqr(x(i)-a) + sqr(y(i)-b) =l= sqr(r);

r.lo = 0;

parameters xmin,ymin,xmax,ymax;
xmin = smin(i, x(i));
ymin = smin(i, y(i));
xmax = smax(i, x(i));
ymax = smax(i, y(i));

* set starting point
a.l = (xmin+xmax)/2;
b.l = (ymin+ymax)/2;
r.l = sqrt( sqr(a.l-xmin) + sqr(b.l-ymin) );

set s scenario s /s1*s27/
Parameter
    s_x(s,i), s_y(s,i);

set dict / s.scenario.''
           x.randvar.s_x
           y.randvar.s_y /;

file emp / '%emp.info%' /; put emp '* problem %gams.i%'; emp.pc=0;
loop(i,
   put / 'jrandvar x("' i.tl:0 '") y("' i.tl:0 '") '
           (1/3):6:2 normal(x(i),1):6:2  normal(y(i),1):6:2
           (1/3):6:2 normal(x(i),1):6:2  normal(y(i),1):6:2
           (1/3):6:2 normal(x(i),1):6:2  normal(y(i),1):6:2;
)
putclose / 'modeltype nlp' / 'stage 2 x y e r'

model m /all/;
solve m using emp minimizing r scenario dict;

* Expanded DE with implicit NA constraints in GAMS using the scenario from previous solve.
equation es(s,i), defobj;
positive variable rs(s);
variable obj;

es(s,i).. sqr(s_x(s,i)-a) + sqr(s_y(s,i)-b) =l= sqr(rs(s));
defobj..  obj =e= sum(s, 1/card(s)*rs(s));

model msi /es, defobj/;

xmin = smin((s,i), s_x(s,i));
ymin = smin((s,i), s_y(S,i));
xmax = smax((s,i), s_x(s,i));
ymax = smax((s,i), s_y(S,i));

* set starting point
a.l = (xmin+xmax)/2;
b.l = (ymin+ymax)/2;
rs.l(s) = sqrt( sqr(a.l-xmin) + sqr(b.l-ymin) );

solve msi using nlp minimizing obj;
abort$(abs(r.l-obj.l)>1e-6) 'EMP and DE (implicit NA) solution differ', r.l, obj.l;

* Expanded DE with explicit NA constraints in GAMS using the scenario from previous solve.
equation ese(s,i), defnaa(s), defnab(s);
positive variable as(s), bs(s);

ese(s,i).. sqr(s_x(s,i)-as(s)) + sqr(s_y(s,i)-bs(s)) =l= sqr(rs(s));
defnaa(s-1).. as('s1') =e= as(s);
defnab(s-1).. bs('s1') =e= bs(s);

model mse /ese, defnaa, defnab, defobj/;

as.l(s) = (xmin+xmax)/2;
bs.l(s) = (ymin+ymax)/2;
rs.l(s) = sqrt( sqr(as.l(s)-xmin) + sqr(bs.l(s)-ymin) );

solve mse using nlp minimizing obj;
abort$(abs(r.l-obj.l)>1e-6) 'EMP and DE (explicit NA) solution differ', r.l, obj.l;