解决 LP 问题

以下部分展示了一个 LP 问题示例，并展示如何解决该问题。问题如下：

尽可能增加 3x + 4y，但需遵循以下限制条件：

1. x + 2y ≤ 14
2. 3x - y ≥ 0
3. x - y ≤ 2

目标函数 3x + 4y 和约束条件均由线性表达式给出，这使得此问题成为线性问题。

这些约束条件定义了可行区域，即下方显示的三角形，包括其内部。

解决 LP 问题的基本步骤

要解决 LP 问题，您的程序应包含以下步骤：

1. 导入线性求解器封装容器，
2. 声明 LP 求解器，
3. 定义变量
4. 定义约束条件，
5. 定义目标，
6. 调用 LP 求解器；
7. 展示解决方案

使用 MPSolver 的解决方案

以下部分展示了一个使用 MPSolver 封装容器和 LP 求解器来解决问题的程序。

注意：如需运行以下程序，您需要安装 OR-Tools。

主要的 OR-Tools 线性优化求解器是 Glop，这是 Google 的内部线性编程求解器。它运行速度快、内存高效且数值稳定。

导入线性求解器封装容器

导入（或包含）OR-Tools 线性求解器封装容器（MIP 求解器和线性求解器的接口），如下所示。

PythonC++JavaC#

from ortools.linear_solver import pywraplp

#include <iostream>
#include <memory>

#include "ortools/linear_solver/linear_solver.h"

import com.google.ortools.Loader;
import com.google.ortools.linearsolver.MPConstraint;
import com.google.ortools.linearsolver.MPObjective;
import com.google.ortools.linearsolver.MPSolver;
import com.google.ortools.linearsolver.MPVariable;

using System;
using Google.OrTools.LinearSolver;

声明 LP 求解器

MPsolver 是多个不同求解器（包括 Glop）的封装容器。以下代码声明了 GLOP 求解器。

PythonC++JavaC#

solver = pywraplp.Solver.CreateSolver("GLOP")
if not solver:
    return

std::unique_ptr<MPSolver> solver(MPSolver::CreateSolver("SCIP"));
if (!solver) {
  LOG(WARNING) << "SCIP solver unavailable.";
  return;
}

MPSolver solver = MPSolver.createSolver("GLOP");

Solver solver = Solver.CreateSolver("GLOP");
if (solver is null)
{
    return;
}

注意：将 GLOP 替换为 PDLP 以使用替代 LP 求解器。如需详细了解如何选择求解器，请参阅高级 LP 求解器；如需了解如何安装第三方求解器，请参阅安装指南。

创建变量

首先，创建变量 x 和 y，它们的值介于 0 到无穷大之间。xx

PythonC++JavaC#

x = solver.NumVar(0, solver.infinity(), "x")
y = solver.NumVar(0, solver.infinity(), "y")

print("Number of variables =", solver.NumVariables())

const double infinity = solver->infinity();
// x and y are non-negative variables.
MPVariable* const x = solver->MakeNumVar(0.0, infinity, "x");
MPVariable* const y = solver->MakeNumVar(0.0, infinity, "y");
LOG(INFO) << "Number of variables = " << solver->NumVariables();

double infinity = java.lang.Double.POSITIVE_INFINITY;
// x and y are continuous non-negative variables.
MPVariable x = solver.makeNumVar(0.0, infinity, "x");
MPVariable y = solver.makeNumVar(0.0, infinity, "y");
System.out.println("Number of variables = " + solver.numVariables());

Variable x = solver.MakeNumVar(0.0, double.PositiveInfinity, "x");
Variable y = solver.MakeNumVar(0.0, double.PositiveInfinity, "y");

Console.WriteLine("Number of variables = " + solver.NumVariables());

定义限制条件

接下来，定义对变量的约束条件。为每个约束条件指定一个唯一的名称（例如 constraint0），然后定义该约束条件的系数。

PythonC++JavaC#

# Constraint 0: x + 2y <= 14.
solver.Add(x + 2 * y <= 14.0)

# Constraint 1: 3x - y >= 0.
solver.Add(3 * x - y >= 0.0)

# Constraint 2: x - y <= 2.
solver.Add(x - y <= 2.0)

print("Number of constraints =", solver.NumConstraints())

// x + 2*y <= 14.
MPConstraint* const c0 = solver->MakeRowConstraint(-infinity, 14.0);
c0->SetCoefficient(x, 1);
c0->SetCoefficient(y, 2);

// 3*x - y >= 0.
MPConstraint* const c1 = solver->MakeRowConstraint(0.0, infinity);
c1->SetCoefficient(x, 3);
c1->SetCoefficient(y, -1);

// x - y <= 2.
MPConstraint* const c2 = solver->MakeRowConstraint(-infinity, 2.0);
c2->SetCoefficient(x, 1);
c2->SetCoefficient(y, -1);
LOG(INFO) << "Number of constraints = " << solver->NumConstraints();

// x + 2*y <= 14.
MPConstraint c0 = solver.makeConstraint(-infinity, 14.0, "c0");
c0.setCoefficient(x, 1);
c0.setCoefficient(y, 2);

// 3*x - y >= 0.
MPConstraint c1 = solver.makeConstraint(0.0, infinity, "c1");
c1.setCoefficient(x, 3);
c1.setCoefficient(y, -1);

// x - y <= 2.
MPConstraint c2 = solver.makeConstraint(-infinity, 2.0, "c2");
c2.setCoefficient(x, 1);
c2.setCoefficient(y, -1);
System.out.println("Number of constraints = " + solver.numConstraints());

// x + 2y <= 14.
solver.Add(x + 2 * y <= 14.0);

// 3x - y >= 0.
solver.Add(3 * x - y >= 0.0);

// x - y <= 2.
solver.Add(x - y <= 2.0);

Console.WriteLine("Number of constraints = " + solver.NumConstraints());

定义目标函数

以下代码定义了目标函数 3x + 4y，并指定这是一个最大化问题。

PythonC++JavaC#

# Objective function: 3x + 4y.
solver.Maximize(3 * x + 4 * y)

// Objective function: 3x + 4y.
MPObjective* const objective = solver->MutableObjective();
objective->SetCoefficient(x, 3);
objective->SetCoefficient(y, 4);
objective->SetMaximization();

// Maximize 3 * x + 4 * y.
MPObjective objective = solver.objective();
objective.setCoefficient(x, 3);
objective.setCoefficient(y, 4);
objective.setMaximization();

// Objective function: 3x + 4y.
solver.Maximize(3 * x + 4 * y);

调用求解器

以下代码会调用求解器。

PythonC++JavaC#

print(f"Solving with {solver.SolverVersion()}")
status = solver.Solve()

const MPSolver::ResultStatus result_status = solver->Solve();
// Check that the problem has an optimal solution.
if (result_status != MPSolver::OPTIMAL) {
  LOG(FATAL) << "The problem does not have an optimal solution!";
}

final MPSolver.ResultStatus resultStatus = solver.solve();

Solver.ResultStatus resultStatus = solver.Solve();

显示解决方案

以下代码将显示解决方案。

PythonC++JavaC#

if status == pywraplp.Solver.OPTIMAL:
    print("Solution:")
    print(f"Objective value = {solver.Objective().Value():0.1f}")
    print(f"x = {x.solution_value():0.1f}")
    print(f"y = {y.solution_value():0.1f}")
else:
    print("The problem does not have an optimal solution.")

LOG(INFO) << "Solution:";
LOG(INFO) << "Optimal objective value = " << objective->Value();
LOG(INFO) << x->name() << " = " << x->solution_value();
LOG(INFO) << y->name() << " = " << y->solution_value();

if (resultStatus == MPSolver.ResultStatus.OPTIMAL) {
  System.out.println("Solution:");
  System.out.println("Objective value = " + objective.value());
  System.out.println("x = " + x.solutionValue());
  System.out.println("y = " + y.solutionValue());
} else {
  System.err.println("The problem does not have an optimal solution!");
}

// Check that the problem has an optimal solution.
if (resultStatus != Solver.ResultStatus.OPTIMAL)
{
    Console.WriteLine("The problem does not have an optimal solution!");
    return;
}
Console.WriteLine("Solution:");
Console.WriteLine("Objective value = " + solver.Objective().Value());
Console.WriteLine("x = " + x.SolutionValue());
Console.WriteLine("y = " + y.SolutionValue());

完整的计划

完整的程序如下所示。

PythonC++JavaC#

from ortools.linear_solver import pywraplp


def LinearProgrammingExample():
    """Linear programming sample."""
    # Instantiate a Glop solver, naming it LinearExample.
    solver = pywraplp.Solver.CreateSolver("GLOP")
    if not solver:
        return

    # Create the two variables and let them take on any non-negative value.
    x = solver.NumVar(0, solver.infinity(), "x")
    y = solver.NumVar(0, solver.infinity(), "y")

    print("Number of variables =", solver.NumVariables())

    # Constraint 0: x + 2y <= 14.
    solver.Add(x + 2 * y <= 14.0)

    # Constraint 1: 3x - y >= 0.
    solver.Add(3 * x - y >= 0.0)

    # Constraint 2: x - y <= 2.
    solver.Add(x - y <= 2.0)

    print("Number of constraints =", solver.NumConstraints())

    # Objective function: 3x + 4y.
    solver.Maximize(3 * x + 4 * y)

    # Solve the system.
    print(f"Solving with {solver.SolverVersion()}")
    status = solver.Solve()

    if status == pywraplp.Solver.OPTIMAL:
        print("Solution:")
        print(f"Objective value = {solver.Objective().Value():0.1f}")
        print(f"x = {x.solution_value():0.1f}")
        print(f"y = {y.solution_value():0.1f}")
    else:
        print("The problem does not have an optimal solution.")

    print("\nAdvanced usage:")
    print(f"Problem solved in {solver.wall_time():d} milliseconds")
    print(f"Problem solved in {solver.iterations():d} iterations")


LinearProgrammingExample()

#include <iostream>
#include <memory>

#include "ortools/linear_solver/linear_solver.h"

namespace operations_research {
void LinearProgrammingExample() {
  std::unique_ptr<MPSolver> solver(MPSolver::CreateSolver("SCIP"));
  if (!solver) {
    LOG(WARNING) << "SCIP solver unavailable.";
    return;
  }

  const double infinity = solver->infinity();
  // x and y are non-negative variables.
  MPVariable* const x = solver->MakeNumVar(0.0, infinity, "x");
  MPVariable* const y = solver->MakeNumVar(0.0, infinity, "y");
  LOG(INFO) << "Number of variables = " << solver->NumVariables();

  // x + 2*y <= 14.
  MPConstraint* const c0 = solver->MakeRowConstraint(-infinity, 14.0);
  c0->SetCoefficient(x, 1);
  c0->SetCoefficient(y, 2);

  // 3*x - y >= 0.
  MPConstraint* const c1 = solver->MakeRowConstraint(0.0, infinity);
  c1->SetCoefficient(x, 3);
  c1->SetCoefficient(y, -1);

  // x - y <= 2.
  MPConstraint* const c2 = solver->MakeRowConstraint(-infinity, 2.0);
  c2->SetCoefficient(x, 1);
  c2->SetCoefficient(y, -1);
  LOG(INFO) << "Number of constraints = " << solver->NumConstraints();

  // Objective function: 3x + 4y.
  MPObjective* const objective = solver->MutableObjective();
  objective->SetCoefficient(x, 3);
  objective->SetCoefficient(y, 4);
  objective->SetMaximization();

  const MPSolver::ResultStatus result_status = solver->Solve();
  // Check that the problem has an optimal solution.
  if (result_status != MPSolver::OPTIMAL) {
    LOG(FATAL) << "The problem does not have an optimal solution!";
  }

  LOG(INFO) << "Solution:";
  LOG(INFO) << "Optimal objective value = " << objective->Value();
  LOG(INFO) << x->name() << " = " << x->solution_value();
  LOG(INFO) << y->name() << " = " << y->solution_value();
}
}  // namespace operations_research

int main(int argc, char** argv) {
  operations_research::LinearProgrammingExample();
  return EXIT_SUCCESS;
}

package com.google.ortools.linearsolver.samples;
import com.google.ortools.Loader;
import com.google.ortools.linearsolver.MPConstraint;
import com.google.ortools.linearsolver.MPObjective;
import com.google.ortools.linearsolver.MPSolver;
import com.google.ortools.linearsolver.MPVariable;

/** Simple linear programming example. */
public final class LinearProgrammingExample {
  public static void main(String[] args) {
    Loader.loadNativeLibraries();
    MPSolver solver = MPSolver.createSolver("GLOP");

    double infinity = java.lang.Double.POSITIVE_INFINITY;
    // x and y are continuous non-negative variables.
    MPVariable x = solver.makeNumVar(0.0, infinity, "x");
    MPVariable y = solver.makeNumVar(0.0, infinity, "y");
    System.out.println("Number of variables = " + solver.numVariables());

    // x + 2*y <= 14.
    MPConstraint c0 = solver.makeConstraint(-infinity, 14.0, "c0");
    c0.setCoefficient(x, 1);
    c0.setCoefficient(y, 2);

    // 3*x - y >= 0.
    MPConstraint c1 = solver.makeConstraint(0.0, infinity, "c1");
    c1.setCoefficient(x, 3);
    c1.setCoefficient(y, -1);

    // x - y <= 2.
    MPConstraint c2 = solver.makeConstraint(-infinity, 2.0, "c2");
    c2.setCoefficient(x, 1);
    c2.setCoefficient(y, -1);
    System.out.println("Number of constraints = " + solver.numConstraints());

    // Maximize 3 * x + 4 * y.
    MPObjective objective = solver.objective();
    objective.setCoefficient(x, 3);
    objective.setCoefficient(y, 4);
    objective.setMaximization();

    final MPSolver.ResultStatus resultStatus = solver.solve();

    if (resultStatus == MPSolver.ResultStatus.OPTIMAL) {
      System.out.println("Solution:");
      System.out.println("Objective value = " + objective.value());
      System.out.println("x = " + x.solutionValue());
      System.out.println("y = " + y.solutionValue());
    } else {
      System.err.println("The problem does not have an optimal solution!");
    }

    System.out.println("\nAdvanced usage:");
    System.out.println("Problem solved in " + solver.wallTime() + " milliseconds");
    System.out.println("Problem solved in " + solver.iterations() + " iterations");
  }

  private LinearProgrammingExample() {}
}

using System;
using Google.OrTools.LinearSolver;

public class LinearProgrammingExample
{
    static void Main()
    {
        Solver solver = Solver.CreateSolver("GLOP");
        if (solver is null)
        {
            return;
        }
        // x and y are continuous non-negative variables.
        Variable x = solver.MakeNumVar(0.0, double.PositiveInfinity, "x");
        Variable y = solver.MakeNumVar(0.0, double.PositiveInfinity, "y");

        Console.WriteLine("Number of variables = " + solver.NumVariables());

        // x + 2y <= 14.
        solver.Add(x + 2 * y <= 14.0);

        // 3x - y >= 0.
        solver.Add(3 * x - y >= 0.0);

        // x - y <= 2.
        solver.Add(x - y <= 2.0);

        Console.WriteLine("Number of constraints = " + solver.NumConstraints());

        // Objective function: 3x + 4y.
        solver.Maximize(3 * x + 4 * y);

        Solver.ResultStatus resultStatus = solver.Solve();

        // Check that the problem has an optimal solution.
        if (resultStatus != Solver.ResultStatus.OPTIMAL)
        {
            Console.WriteLine("The problem does not have an optimal solution!");
            return;
        }
        Console.WriteLine("Solution:");
        Console.WriteLine("Objective value = " + solver.Objective().Value());
        Console.WriteLine("x = " + x.SolutionValue());
        Console.WriteLine("y = " + y.SolutionValue());

        Console.WriteLine("\nAdvanced usage:");
        Console.WriteLine("Problem solved in " + solver.WallTime() + " milliseconds");
        Console.WriteLine("Problem solved in " + solver.Iterations() + " iterations");
    }
}

最佳解决方案

程序会返回问题的最佳解决方案，如下所示。

Number of variables = 2
Number of constraints = 3
Solution:
x = 6.0
y = 4.0
Optimal objective value = 34.0

下图显示了该解决方案：

绿色虚线通过将目标函数设置为等于其最佳值 34 来定义。任何方程式为 3x + 4y = c 的直线都平行于虚线，且 34 是该直线与可行区域相交的 c 的最大值。

如需详细了解如何解决线性优化问题，请参阅高级 LP 求解。