public void ExecuteMetaheuristic(ref Solution bestSolution)
{
int bestIteration = 0;
Solution partialSolution = GenericCopier <Solution> .DeepCopy(bestSolution);
int control_Diversify = 0;
SummaryDetails summary = new SummaryDetails();
Splash.SplashGlobalData.SetSplashData(Constants.SPLASH_BEST_SOLUTION, bestSolution.Fitness);
double pertCost = 0, vnsCost = 0;
int iterations = 0;
//Guardar coste inicial.
summary.ILSSummary.InitCost = bestSolution.Parc_Cost;
for (int i = 0; iterations < _ilsSettings.MaxILSIterations && i < _ilsSettings.MaxILSNoImprovement; i++)
{
//Control de intensificación-diversificación.
if (i % _DIVERSIFY_MAX == 0)
{
_diversify--;
}
_diversify = _diversify < 3 ? 3 : _diversify;
if (i % 100 == 0)
{
_diversify = _DIVERSIFY_MAX;
}
//Guardar valores estadísticos.
var initCost = partialSolution.Parc_Cost;
var initFitness = partialSolution.Fitness;
Splash.SplashGlobalData.SetSplashData(Constants.SPLASH_ILS_TOTAL_ITERATIONS, iterations + 1);
//Apply VNS local search.
_vns.ExecuteMetaheuristic(ref partialSolution);
//Validator.ValidateSolution(ref partialSolution, _problem);
bool feasible = partialSolution.Feasible;
//Guardar valores estadísticos.
vnsCost = Costs.ParcialCostSolution(partialSolution, _problem); // partialSolution.Parc_Cost;
var vnsFitness = partialSolution.Fitness;
if (vnsFitness < initFitness)
{
summary.ILSSummary.TotalImpPrevious++; //Incrementar el número de veces que se mejora la solución previa (perturbada).
}
if (bestSolution.Feasible)
{
if (partialSolution.Feasible && (Math.Round(partialSolution.Fitness, 2) < Math.Round(bestSolution.Fitness, 2)))
{
bestSolution = GenericCopier <Solution> .DeepCopy(partialSolution);
//Actualizar la mejor solución con los nuevos parámetros.
control_Diversify = 0;
bestIteration = i;
i = 0;
_diversify = _DIVERSIFY_MAX;
summary.ILSSummary.BestIteration = iterations + 1; //Guardar mejor iteración encontrada.
summary.ILSSummary.TotalImpBest++; //Incrementar el número de veces que se mejora la mejor solución.
Singleton.Instance.UpdatePenaltyTerms(partialSolution);
Splash.SplashGlobalData.SetSplashData(Constants.SPLASH_ILS_IMPROVEMENTS, summary.ILSSummary.TotalImpBest);
Splash.SplashGlobalData.SetSplashData(Constants.SPLASH_BEST_SOLUTION, Math.Round(bestSolution.Fitness, 2));
}
else
{
control_Diversify++;
}
}
else
{
if (partialSolution.Feasible || Math.Round(partialSolution.Fitness, 2) < Math.Round(bestSolution.Fitness, 2))
{
if (partialSolution.Feasible && summary.ILSSummary.FirstItFeasible == null)
{
summary.ILSSummary.FirstItFeasible = iterations + 1; //Guardar primera iteración factible.
}
bestSolution = GenericCopier <Solution> .DeepCopy(partialSolution);
control_Diversify = 0;
i = 0;
_diversify = _DIVERSIFY_MAX;
summary.ILSSummary.BestIteration = iterations + 1; //Guardar mejor iteración encontrada.
summary.ILSSummary.TotalImpBest++; //Incrementar el número de veces que se mejora la mejor solución.
Singleton.Instance.UpdatePenaltyTerms(partialSolution);
//bestSolution.Fitness = Costs.EvaluationSolution(bestSolution, _problem);
Splash.SplashGlobalData.SetSplashData(Constants.SPLASH_ILS_IMPROVEMENTS, summary.ILSSummary.TotalImpBest);
Splash.SplashGlobalData.SetSplashData(Constants.SPLASH_BEST_SOLUTION, bestSolution.Fitness);
}
else
{
control_Diversify++;
}
}
if (control_Diversify > _diversify)
{
partialSolution = GenericCopier <Solution> .DeepCopy(bestSolution);
control_Diversify = 0;
}
var perturbation = _perturbations.ExecutePerturbation(ref partialSolution);
//Guardar valores estadísticos.
pertCost = partialSolution.Parc_Cost;
var pertFitness = partialSolution.Fitness;
//Añadir resumen de la iteración.
summary.VNSOperators.Add(new VNSOperators(iterations, _vns.ShiftInter, 0, _vns.WRI_IntraOP, _vns.IntraRouteInsertionOP, _vns.IntraSwap));
summary.ILSEvolution.Add(new ILSEvolution(iterations, Math.Round(initCost, 2), Math.Round(initFitness, 2), Math.Round(vnsCost, 2), Math.Round(vnsFitness, 2), perturbation, Math.Round(pertFitness, 2), Math.Round(pertCost, 2), feasible));
Splash.SplashGlobalData.SetSplashData(Constants.SPLASH_PROGRESS, Convert.ToDouble(Splash.SplashGlobalData.GetSplashData <double>(Constants.SPLASH_PROGRESS)) + _coef_Updater_Splash);
//Guardar cambios de cada operador.
iterations++;
}
//Recalcular con el proceso de 8 pasos y actualizar el coste.
foreach (var vehicle in bestSolution.Vehicles)
{
DARPAlgorithms.EightStepsEvaluationProcedure(ref bestSolution, vehicle, _problem);
}
bestSolution.TotalDuration = bestSolution.Vehicles.Sum(t => t.VehicleDuration);
bestSolution.TimeWindowViolation = bestSolution.Vehicles.Sum(t => t.TotalTimeWindowViolation);
bestSolution.RideTimeViolation = bestSolution.Vehicles.Sum(r => r.TotalRideTimeViolation);
bestSolution.DurationViolation = bestSolution.Vehicles.Sum(d => d.TotalDurationViolation);
bestSolution.LoadViolation = bestSolution.Vehicles.Sum(l => l.TotalLoadViolation);
bestSolution.TotalWaitingTime = bestSolution.Vehicles.Sum(t => t.TotalWaitingTime);
DARPAlgorithms.UpdateConstraintsSolution(ref bestSolution, _problem);
bestSolution.Fitness = Costs.EvaluationSolution(bestSolution, _problem);
bestSolution.Parc_Cost = Costs.ParcialCostSolution(bestSolution, _problem);
summary.ILSSummary.TotalIterations = iterations + 1;
summary.ILSSummary.FinalCost = bestSolution.Parc_Cost;
bestSolution.SummaryDetails = summary;
}
