Accessing intermediate solutions

As CP Optimizer searched for an optimal solution to an optimization problem, the search will generally encounter a sequence of solutions that improve the objective function.

As the CP Optimizer engine solves an optimization problem, a sequence of solutions that improve the objective function are produced until an optimal solution is found. In some cases, you may want to have access to this sequence of intermediate solutions. CP Optimizer provides a simple interface that provides this access via member functions of the optimizer object.

To prepare the optimizer for search, you call the member function startNewSearch, a member function of the optimizer object. To instruct the optimizer to find a solution, you call the method next. This method returns true (IloTrue in the C++ API) if the optimizer finds a solution (not necessarily an optimal one) and false (IloFalse in the C++ API) if the problem is infeasible. To find the next solution, you again call the method next. If the method returns true, then the optimizer has found a new solution with a strictly better objective value than the previous one. If the optimizer does not find another solution, the value returned by this method is false. When you have finished searching, you can call the method end to free the memory and reset the state of the optimizer. A typical code using these methods of the optimizer object in the C++ API would be similar to:

    IloCP cp(model);
    cp.setParameter(IloCP::LogVerbosity, IloCP::Quiet);
    cp.startNewSearch();
    while(cp.next()){
      cp.out() << "objective value = " << cp.getObjValue()
               << ", x = " << cp.getValue(x)
               << ", y = " << cp.getValue(y)
               << ", z = " << cp.getValue(z) << std::endl;
    }
    cp.end();

Running this code produces the output:

objective value = 5, x = 4, y = 0, z = 1
objective value = 4, x = 0, y = 3, z = 1

This run finds two solutions, one with an objective value of 5 and then one with an objective value of 4, which is the optimum.

In the C++ API of CP Optimizer, you use the class IloCP and the methods IloCP::startNewSearch and IloCP::next.

In the Java™ API of CP Optimizer, you use the class IloCP and the methods IloCP.startNewSearch and IloCP.next.

In the C# API of CP Optimizer, you use the class CP and the methods CP.StartNewSearch and CP.Next.