/**
* Constraint modeling the Traveling Salesman Problem
*
* @param GRAPHVAR graph variable representing a Hamiltonian cycle
* @param COSTVAR variable representing the cost of the cycle
* @param EDGE_COSTS cost matrix (should be symmetric)
* @param LAGR_MODE use the Lagrangian relaxation of the tsp described by Held and Karp {0:no
* Lagrangian relaxation, 1:Lagrangian relaxation (since root node), 2:Lagrangian relaxation
* but wait a first solution before running it}
* @return a tsp constraint
*/
public static Constraint tsp(
IUndirectedGraphVar GRAPHVAR, IntVar COSTVAR, int[][] EDGE_COSTS, int LAGR_MODE) {
Propagator[] props =
ArrayUtils.append(
hamiltonian_cycle(GRAPHVAR).getPropagators(),
new Propagator[] {new PropCycleCostSimple(GRAPHVAR, COSTVAR, EDGE_COSTS)});
if (LAGR_MODE > 0) {
PropLagr_OneTree hk = new PropLagr_OneTree(GRAPHVAR, COSTVAR, EDGE_COSTS);
hk.waitFirstSolution(LAGR_MODE == 2);
props = ArrayUtils.append(props, new Propagator[] {hk});
}
return new Constraint("TSP", props);
}
/**
* AtLeastNValues Propagator (similar to SoftAllDiff) The number of distinct values in vars is at
* least nValues Performs Generalized Arc Consistency based on Maximum Bipartite Matching The
* worst case time complexity is O(nm) but this is very pessimistic In practice it is more like
* O(m) where m is the number of variable-value pairs
*
* @param variables array of integer variables
* @param nValues integer variable
*/
public PropAtLeastNValues_AC(IntVar[] variables, IntVar nValues) {
super(ArrayUtils.append(variables, new IntVar[] {nValues}), PropagatorPriority.QUADRATIC, true);
this.idms = new IIntDeltaMonitor[this.vars.length];
for (int i = 0; i < this.vars.length; i++) {
idms[i] = this.vars[i].monitorDelta(this);
}
n = variables.length;
map = new TIntIntHashMap();
IntVar v;
int ub;
int idx = n;
for (int i = 0; i < n; i++) {
v = vars[i];
ub = v.getUB();
for (int j = v.getLB(); j <= ub; j = v.nextValue(j)) {
if (!map.containsKey(j)) {
map.put(j, idx);
idx++;
}
}
}
n2 = idx;
fifo = new int[n2];
digraph = new DirectedGraph(model, n2 + 2, SetType.LINKED_LIST, false);
free = new BitSet(n2);
remProc = new DirectedRemProc();
father = new int[n2];
in = new BitSet(n2);
SCCfinder = new StrongConnectivityFinder(digraph);
}
public PropBoolMax(BoolVar[] variables, BoolVar maxVar) {
super(ArrayUtils.append(variables, new BoolVar[] {maxVar}), PropagatorPriority.UNARY, true);
n = variables.length;
x1 = -1;
x2 = -1;
assert n > 0;
}
public PropKnapsack(
IntVar[] itemOccurence, IntVar capacity, IntVar power, int[] weight, int[] energy) {
super(
ArrayUtils.append(itemOccurence, new IntVar[] {capacity, power}),
PropagatorPriority.LINEAR,
false);
this.weigth = weight;
this.energy = energy;
this.n = itemOccurence.length;
this.capacity = vars[n];
this.power = vars[n + 1];
this.order = new int[n];
this.ratio = new double[n];
for (int i = 0; i < n; i++) {
ratio[i] = weight[i] == 0 ? Double.MAX_VALUE : ((double) (energy[i]) / (double) (weight[i]));
}
this.order = ArrayUtils.array(0, n - 1);
ArraySort sorter = new ArraySort(n, false, true);
sorter.sort(
order,
n,
(i1, i2) -> {
return Double.compare(ratio[i2], ratio[i1]);
});
}
public PropIntEqReal(IntVar[] intVars, RealVar[] realVars, double epsilon) {
super(ArrayUtils.append(intVars, realVars), PropagatorPriority.LINEAR, false);
this.n = intVars.length;
this.intVars = intVars;
this.realVars = realVars;
this.epsilon = epsilon;
assert n == realVars.length;
}
/**
* Creates a degree-constrained minimum spanning tree constraint : GRAPH is a spanning tree of
* cost COSTVAR and each vertex degree is constrained
*
* <p>BEWARE : assumes the channeling between GRAPH and DEGREES is already done
*
* @param GRAPH an undirected graph variable
* @param DEGREES the degree of every vertex
* @param COSTVAR variable representing the cost of the mst
* @param EDGE_COSTS cost matrix (should be symmetric)
* @param LAGR_MODE use the Lagrangian relaxation of the dcmst {0:no Lagrangian relaxation,
* 1:Lagrangian relaxation (since root node), 2:Lagrangian relaxation but wait a first
* solution before running it}
* @return a degree-constrained minimum spanning tree constraint
*/
public static Constraint dcmst(
IUndirectedGraphVar GRAPH,
IntVar[] DEGREES,
IntVar COSTVAR,
int[][] EDGE_COSTS,
int LAGR_MODE) {
Propagator[] props =
ArrayUtils.append(
tree(GRAPH).getPropagators(),
new Propagator[] {
new PropTreeCostSimple(GRAPH, COSTVAR, EDGE_COSTS),
new PropMaxDegVarTree(GRAPH, DEGREES)
});
if (LAGR_MODE > 0) {
PropLagr_DCMST_generic hk =
new PropLagr_DCMST_generic(GRAPH, COSTVAR, DEGREES, EDGE_COSTS, LAGR_MODE == 2);
props = ArrayUtils.append(props, new Propagator[] {hk});
}
return new Constraint("dcmst", props);
}
/** Inverse set propagator x in sets[y-offSet1] <=> y in inverses[x-offSet2] */
public PropInverse(SetVar[] sets, SetVar[] invsets, int offSet1, int offSet2) {
super(ArrayUtils.append(sets, invsets), PropagatorPriority.LINEAR, true);
n = sets.length;
n2 = invsets.length;
this.offSet1 = offSet1;
this.offSet2 = offSet2;
this.sets = Arrays.copyOfRange(vars, 0, sets.length);
this.invsets = Arrays.copyOfRange(vars, sets.length, vars.length);
// delta monitors
sdm = new ISetDeltaMonitor[n + n2];
for (int i = 0; i < n + n2; i++) {
sdm[i] = this.vars[i].monitorDelta(this);
}
elementForced = element -> toFilter[element - offSet].addToKernel(idx, aCause);
elementRemoved = element -> toFilter[element - offSet].removeFromEnvelope(idx, aCause);
}
public PropIntValuesUnion(IntVar[] X, IntVar union) {
super(ArrayUtils.append(new IntVar[] {union}, X), PropagatorPriority.LINEAR, false);
}
@Override
public void configureSearch() {
solver.set(IntStrategyFactory.minDom_LB(ArrayUtils.append(rows)));
}