SONET problem in Picat.
From the EssencePrime model in the Minion Translator examples:
The SONET problem is a network design problem: set up a network between
n nodes, where only certain nodes require a connection.
Nodes are connected by putting them on a ring, where all nodes
on a ring can communicate. Putting a node on a ring requires a so-called
ADM, and each ring has a capacity of nodes, i.e. ADMs. There is a certain
amount of rings, r, that is available. The objective is to set up a network
by using a minimal amount of ADMs.
About the problem model
The problem model has the amount of rings ('r'), amount of nodes('n'),
the 'demand' (which nodes require communication) and node-capacity of each
ring ('capacity_nodes') as parameters.
The assignement of nodes to rings is modelled by a 2-dimensional matrix 'rings',
indexed by the amnount of rings and nodes. The matrix-domain is boolean:
If the node in column j is assigned to the ring in row i, then rings[i,j] = 1
and 0 otherwise. So all the '1's in the matrix 'rings' stand for an ADM.
Hence the objective is to minimise the sum over all columns and rows of matrix
Model created by Hakan Kjellerstrand, email@example.com
See also my Picat page: http://www.hakank.org/picat/
% Licenced under CC-BY-4.0 : http://creativecommons.org/licenses/by/4.0/
main => go.
R = 4,
N = 5,
Demand = [[0,1,0,1,0],
CapacityNodes = [3,2,2,1],
% decision variables
Rings = new_array(R,N).array_matrix_to_list_matrix(),
Vars = vars(Rings),
Vars :: 0..1,
% to optimize
Z #= sum([(Rings[Ring,Client]) : Ring in 1..R, Client in 1..N]),
% if there is a demand between 2 nodes, then there has to exist
% a ring, on which they are both installed
foreach(Client1 in 1..N, Client2 in Client1+1..N)
if Demand[Client1,Client2] == 1 then
R1 + R2 #>= 2
% capacity of each ring must not be exceeded
foreach(Ring in 1..R)
sum([Rings[Ring,Client] : Client in 1..N]) #=< CapacityNodes[Ring]
% Z #= 7, % for showing all 6 optimal solutions
foreach(RR in Rings) writeln(RR) end,
matrix_element(X, I, J, Val) =>
nth(I, X, Row),
element(J, Row, Val).
% matrix_element(X, I, J, Val) =>
% freeze(I, (element(I, X, Row),freeze(J,element(J,Row,Val)))).