Version:

mip05.gms : Maximum queens chess problem

**Description**

This model finds all possible ways to arrange eight queens on a chess board in such a way that no two queens check each other. Using solves within a loop, successive solutions are cut off until the model becomes infeasible. At this point, all the solutions have been enumerated. This problem has a long history. In 1850 it was investigated by C.F. Gauss, but he was unable to solve it completely. There are 92 possible solutions on a standard 8x8 chessboard. Alternate NxN chessboards can easily be used, see numSols below. Dudeney, H E, Amusements in Mathematics. Dover, New York, 1970. Beauvais, J, Solving the Maximum Queens Chess Problem with OSL. IBM Kingston, EKKNEWS 2 (1991). GAMS Model Library, queens.103 Contributor: Steve Dirkse, Jan 2012

**Small Model of Type : ** MIP

**Category :** GAMS Test library

**Main file : ** mip05.gms

```
$Title Maximum queens chess problem (MIP05,SEQ=549)
$Eolcom !
$ontext
This model finds all possible ways to arrange eight queens on a chess
board in such a way that no two queens check each other.
Using solves within a loop, successive solutions are cut off until
the model becomes infeasible. At this point, all the solutions
have been enumerated. This problem has a long history. In 1850
it was investigated by C.F. Gauss, but he was unable to solve it
completely. There are 92 possible solutions on a standard 8x8 chessboard.
Alternate NxN chessboards can easily be used, see numSols below.
Dudeney, H E, Amusements in Mathematics. Dover, New York, 1970.
Beauvais, J, Solving the Maximum Queens Chess Problem with OSL. IBM
Kingston, EKKNEWS 2 (1991).
GAMS Model Library, queens.103
Contributor: Steve Dirkse, Jan 2012
$offtext
* to run without cut generation use --SKIPCUTS=1 on the command line
* this will just solve the problem without cuts and by default will
* only return one solution
$if not set SKIPCUTS $set SKIPCUTS 0
$ifthen %DEMOSIZE% == 1
set i 'size of chess board' / 1*6 /;
$else
set i 'size of chess board' / 1*8 /;
$endif
$eval NS 2 * card(i) - 3
sets
s 'diagonal offsets' / 1*%NS% / ! 2i-3 diagonals
nn 'max number of solutions' / s1*s1000 /
n(nn) 'solutions found'
dim 'known dimensions' / d1 * d10 /
;
alias (i,j);
parameters
sh(s) 'shift values for diagonals'
rev(i) 'reverse order'
sols(nn,i,j) 'previously found solutions for cut generation'
numSols(dim) /
d4 2
d5 10
d6 4
d7 40
d8 92
d9 352
d10 724
/
;
scalars
nI 'size of chessboard'
nSols 'number of solutions found'
;
sh(s) = ord(s) - card(i) + 1 ;
rev(i) = card(i) + 1 - 2*ord(i);
Display sh, rev;
binary variable x(i,j) 'square is occupied by a queen'
variable tot 'count of squares occupied by queens'
equations
a(i) 'no two queens may be in the same rank'
b(j) 'no two queens may be in the same file'
c(s) 'no two queens may be in the same diagonal (forward)'
d(s) 'no two queens may be in the same diagonal (backward)'
obj 'objective definition'
cut(nn) 'known solutions to be eliminated'
;
a(i).. sum(j, x(i,j)) =e= 1;
b(j).. sum(i, x(i,j)) =e= 1;
c(s).. sum(i, x(i,i+sh(s))) =l= 1;
d(s).. sum(i, x(i,i+(rev(i)+sh(s)))) =l= 1;
cut(n).. sum{(i,j), sols(n,i,j)*x(i,j)} =l= card(i)-1;
obj.. tot =e= sum{(i,j), x(i,j)};
Model queens 'model with cuts' / all /;
n(nn) = no; ! clear set of cuts
sols(nn,i,j) = 0;
option optcr=0;
solve queens maximizing tot using mip;
$if %SKIPCUTS% == 1 $exit
option limrow=0, limcol=0, solprint=off;
option solveLink = %SOLVELINK.loadLibrary%;
loop{nn$(queens.solvestat=%solvestat.NormalCompletion% and
queens.modelstat=%modelstat.Optimal%),
n(nn) = yes; ! add element to set
sols(nn,i,j) = round(x.l(i,j));
Solve queens maximizing tot using mip;
};
abort$[card(n) eq card(nn)] 'set nn too small';
nI = card(i);
nSols = card(n);
display nI, nSols;
abort$[nSols <> sum{dim$(ord(dim) eq nI), numSols(dim)}] 'bad nSols', nSols, numSols, nI;
execute_unload 'queensSolsCuts', n, I, sols;
```