Try our new documentation site (beta).
Filter Content By
Version
Text Search
${sidebar_list_label} - Back
Filter by Language
Diet.java
/* Copyright 2020, Gurobi Optimization, LLC */ /* Solve the classic diet model, showing how to add constraints to an existing model. */ import gurobi.*; public class Diet { public static void main(String[] args) { try { // Nutrition guidelines, based on // USDA Dietary Guidelines for Americans, 2005 // http://www.health.gov/DietaryGuidelines/dga2005/ String Categories[] = new String[] { "calories", "protein", "fat", "sodium" }; int nCategories = Categories.length; double minNutrition[] = new double[] { 1800, 91, 0, 0 }; double maxNutrition[] = new double[] { 2200, GRB.INFINITY, 65, 1779 }; // Set of foods String Foods[] = new String[] { "hamburger", "chicken", "hot dog", "fries", "macaroni", "pizza", "salad", "milk", "ice cream" }; int nFoods = Foods.length; double cost[] = new double[] { 2.49, 2.89, 1.50, 1.89, 2.09, 1.99, 2.49, 0.89, 1.59 }; // Nutrition values for the foods double nutritionValues[][] = new double[][] { { 410, 24, 26, 730 }, // hamburger { 420, 32, 10, 1190 }, // chicken { 560, 20, 32, 1800 }, // hot dog { 380, 4, 19, 270 }, // fries { 320, 12, 10, 930 }, // macaroni { 320, 15, 12, 820 }, // pizza { 320, 31, 12, 1230 }, // salad { 100, 8, 2.5, 125 }, // milk { 330, 8, 10, 180 } // ice cream }; // Model GRBEnv env = new GRBEnv(); GRBModel model = new GRBModel(env); model.set(GRB.StringAttr.ModelName, "diet"); // Create decision variables for the nutrition information, // which we limit via bounds GRBVar[] nutrition = new GRBVar[nCategories]; for (int i = 0; i < nCategories; ++i) { nutrition[i] = model.addVar(minNutrition[i], maxNutrition[i], 0, GRB.CONTINUOUS, Categories[i]); } // Create decision variables for the foods to buy GRBVar[] buy = new GRBVar[nFoods]; for (int j = 0; j < nFoods; ++j) { buy[j] = model.addVar(0, GRB.INFINITY, cost[j], GRB.CONTINUOUS, Foods[j]); } // The objective is to minimize the costs model.set(GRB.IntAttr.ModelSense, GRB.MINIMIZE); // Nutrition constraints for (int i = 0; i < nCategories; ++i) { GRBLinExpr ntot = new GRBLinExpr(); for (int j = 0; j < nFoods; ++j) { ntot.addTerm(nutritionValues[j][i], buy[j]); } model.addConstr(ntot, GRB.EQUAL, nutrition[i], Categories[i]); } // Solve model.optimize(); printSolution(model, buy, nutrition); System.out.println("JSON solution:" + model.getJSONSolution()); System.out.println("\nAdding constraint: at most 6 servings of dairy"); GRBLinExpr lhs = new GRBLinExpr(); lhs.addTerm(1.0, buy[7]); lhs.addTerm(1.0, buy[8]); model.addConstr(lhs, GRB.LESS_EQUAL, 6.0, "limit_dairy"); // Solve model.optimize(); printSolution(model, buy, nutrition); System.out.println("JSON solution:" + model.getJSONSolution()); // Dispose of model and environment model.dispose(); env.dispose(); } catch (GRBException e) { System.out.println("Error code: " + e.getErrorCode() + ". " + e.getMessage()); } } private static void printSolution(GRBModel model, GRBVar[] buy, GRBVar[] nutrition) throws GRBException { if (model.get(GRB.IntAttr.Status) == GRB.Status.OPTIMAL) { System.out.println("\nCost: " + model.get(GRB.DoubleAttr.ObjVal)); System.out.println("\nBuy:"); for (int j = 0; j < buy.length; ++j) { if (buy[j].get(GRB.DoubleAttr.X) > 0.0001) { System.out.println(buy[j].get(GRB.StringAttr.VarName) + " " + buy[j].get(GRB.DoubleAttr.X)); } } System.out.println("\nNutrition:"); for (int i = 0; i < nutrition.length; ++i) { System.out.println(nutrition[i].get(GRB.StringAttr.VarName) + " " + nutrition[i].get(GRB.DoubleAttr.X)); } } else { System.out.println("No solution"); } } }