Batch Scheduling Problem¶

Principles learned¶

  • Use the distinct operator

  • Enforce precedence of some tasks

  • Use interval decision variables to schedule batches of tasks

  • Assign tasks to batches using sets and partitioning

Problem¶

In the batch scheduling problem, a set of tasks must be processed in batches on a set of available resources. The set of tasks must be sorted into batches. Each batch must consist of tasks with the same type and the same task duration. Each batch of tasks must be processed on one resource. For this reason, all tasks in a batch must have the same resource type. Each resource can only process one batch of tasks at a time. The number of tasks in a batch cannot exceed the maximum capacity of the resource it is assigned to. Additionally, some tasks have to be processed before others.

The goal is to find the intervals of time each task should be processed on a resource that minimizes the makespan: the time when all jobs have been processed. This is accomplished by assigned tasks to batches and scheduling batches across the resources.

Download the example


Data¶

The format of the data files is as follows:

  • First line: number of tasks, number of resources.

  • Second line: for each resource, the maximum capacity

  • For each task: task resource assignment, task type, task duration, number of tasks that must succeed this task, task numbers that must succeed.

Program¶

We use set decision variables to determine the tasks assigned to each batch. For each resource, the largest number of possible batches is equal to the number of tasks to be processed on that resource. Therefore, we have a resource specific collection of sets (of this maximum size) to model the batches. Each set determines the content of the batch (i.e., which tasks are assigned to each batch). To make sure each task is assigned to exactly one batch, we use the partition operator on each resource’s collection of sets.

The tasks in a batch must have the same type, so we use the distinct operator coupled with the count operator. The distinct operator returns an unordered set of distinct values among the input collection. We use this within the count operator to constrain the number of task types to be the same.

We cannot exceed the maximum capacity of a resource. We again use the count operator to constrain the size of a batch associated with a resource not to exceed the capacity of the resource.

In addition to the set decisions representing the tasks assigned to each batch, we use interval decision variables to model the time ranges of each batch of tasks on a resource. The length of the interval is constrained to match the length of the tasks within the batch (recall that each task in a batch must have the same length).

The disjunctive resource constraint ensures each resource can only process one batch at a time. Given a sequence of batches, the interval corresponding to batch n must start after the interval corresponding to batch (n-1).

We use the find operator to map a task interval to the resource and batch interval that contains the task. We use this mapping to enforce precedence constraints on the tasks that must be processed before other tasks.

The makespan to minimize is the time when all activities have been processed.

Execution:
hexaly batch_scheduling.hxm inFileName=instances/5_2-000.nfb [hxTimeLimit=] [solFileName=]
use io;

/* Read instance data. */
function input() {
    local usage = "Usage: hexaly batch_scheduling.hxm "
            + "inFileName=inputFile [solFileName=outputFile] [hxTimeLimit=timeLimit]";

    if (inFileName == nil) throw usage;
    local inFile = io.openRead(inFileName);

    // Number of tasks
    nbTasks = inFile.readInt();
    // Number of resources
    nbResources = inFile.readInt();
    // Maximum capacity of each resource
    capacity[r in 0...nbResources] = inFile.readInt();

    // Number of tasks assigned to each resource
    nbTasksPerResource[r in 0...nbResources] = 0;
    // Types of tasks assigned to each resource
    typesInResource[r in 0...nbResources] = {};
    // Tasks assigned to each resource
    tasksInResource[r in 0...nbResources] = {};

    for [i in 0...nbTasks] {

        // Type of task i
        type[i] = inFile.readInt();
        // Resource required for task i
        resource[i] = inFile.readInt();

        // Index of task i on resource[i]
        taskIndexInResource[i] = nbTasksPerResource[resource[i]];
        // Map from index of task i on resource[i] to task i type
        typesInResource[resource[i]][taskIndexInResource[i]] = type[i];
        // Map from index of task i on resource[i] to task i id
        tasksInResource[resource[i]][taskIndexInResource[i]] = i;
        // Increment number of tasks required by this resource
        nbTasksPerResource[resource[i]] = nbTasksPerResource[resource[i]] + 1;

        // Duration of task i
        duration[i] = inFile.readInt();

        // Number of tasks that must succeed task i
        nbSuccessors[i] = inFile.readInt();
        // Task id that must succeed task i
        successors[i][s in 0...nbSuccessors[i]] = inFile.readInt();
    }

    // Longest possible time horizon
    timeHorizon = sum[t in 0...nbTasks](duration[t]);

    inFile.close();
}

/* Declare the optimization model */
function model() {

    // For each resource, the contents of each batch of tasks performed
    batchContent[r in 0...nbResources][b in 0...nbTasksPerResource[r]] <- set(nbTasksPerResource[r]);

    // All tasks are assigned to a batch
    for [r in 0...nbResources] {
        constraint partition(batchContent[r]);
    }

    // Each batch must consist of tasks with the same type 
    for [r in 0...nbResources][batch in batchContent[r]] {
        constraint count(distinct(batch, i => typesInResource[r][i])) <= 1;
    }

    // Each batch cannot exceed the maximum capacity of the resource
    for [r in 0...nbResources][batch in batchContent[r]] {
        constraint count(batch) <= capacity[r];
    }

    // Interval decisions: time range of each batch of tasks
    batchInterval[r in 0...nbResources][n in 0...nbTasksPerResource[r]] <- interval(0, timeHorizon);

    // Non-overlap of batch intervals on the same resource
    for [r in 0...nbResources][b in 1...nbTasksPerResource[r]] {
        constraint batchInterval[r][b-1] < batchInterval[r][b];
    }

    // Interval decisions: time range of each task
    for [t in 0...nbTasks] {
        // Retrieve the batch index and resource for this task
        local r = resource[t];
        local b <- find(batchContent[r], taskIndexInResource[t]);
        // Task interval interval associated with task t
        taskInterval[t] <- batchInterval[r][b];
    }

    // Task durations
    for [t in 0...nbTasks] {
        constraint length(taskInterval[t]) == duration[t];
    }

    // Precedence constraints between tasks
    for [t in 0...nbTasks][s in successors[t]] {
        constraint taskInterval[t] < taskInterval[s];
    }

    // Makespan: end of the last task
    makespan <- max[t in 0...nbTasks](end(taskInterval[t]));
    minimize makespan;
}

/* Parameterize the solver */
function param() {
    if (hxTimeLimit == nil) hxTimeLimit = 15;
}

/* Write the solution in a file with the following format:
 *  - makespan
    - machine number
    - preceeding lines are the ordered intervals of the tasks 
      for the corresponding machine number */
function output() {
    if (outFileName != nil) {
        outFile = io.openWrite(outFileName);
        outFile.println(makespan.value);
        for [r in 0...nbResources] {
            outFile.println(r);
            for [b in 0...nbTasksPerResource[r]] {
                local t = tasksInResource[r][b];
                outFile.println(t + " " + taskInterval[t].value.start + " " + taskInterval[t].value.end);
            }
        }
        outFile.close();
    }
}
Execution (Windows)
set PYTHONPATH=%HX_HOME%\bin\python
python batch_scheduling.py instances\5_2-000.nfb
Execution (Linux)
export PYTHONPATH=/opt/hexaly_13_0/bin/python
python batch_scheduling.py instances/5_2-000.nfb
import hexaly.optimizer
import sys

def read_instance(filename):
    # The import files follow the "Taillard" format
    with open(filename, 'r') as f:
        lines = f.readlines()

    first_line = lines[0].split()
    # Number of tasks
    nb_tasks = int(first_line[0])
    # Number of resources
    nb_resources = int(first_line[1])

    second_line = lines[1].split()
    # Capacity of each resource
    capacity = [int(l) for l in second_line]

    # Remaining lines contain task information at each row
    nb_tasks_per_resource = [0 for _ in range(nb_resources)]
    types_in_resource = [[] for _ in range(nb_resources)]
    tasks_in_resource = [[] for _ in range(nb_resources)]
    task_index_in_resource = []
    types, resources, duration, nb_successors = [], [], [], []
    successors = [[] for _ in range(nb_tasks)]

    for i in range(nb_tasks):

        # Extract dataset line related to this task
        task_line = i + 2
        task_information = lines[task_line].split()
        
        # Type of task i
        types.append(int(task_information[0]))
        # Resource required for task i
        resources.append(int(task_information[1]))
        
        # Index of task i on resource[i]
        task_index_in_resource.append(nb_tasks_per_resource[resources[i]])
        # Map from name of task i on resource[i] to task i type
        types_in_resource[resources[i]].append(types[i])
        # Map from name of task i on resource[i] to task i
        tasks_in_resource[resources[i]].append(i)
        # Increment number of tasks required by this resource
        nb_tasks_per_resource[resources[i]] += 1
        
        # Task duration
        duration.append(int(task_information[2]))

        # Number of successors of this task
        nb_successors.append(int(task_information[3]))
        # Tasks that must succeed current task
        for succeeding_task in task_information[4:]:
            successors[i].append(int(succeeding_task))

    # Trivial time horizon
    time_horizon = sum(duration[t] for t in range(nb_tasks))

    return nb_tasks, nb_resources, capacity, types, resources, duration, \
            nb_successors, successors, nb_tasks_per_resource, \
              task_index_in_resource, types_in_resource, tasks_in_resource, \
                nb_tasks_per_resource, time_horizon


def main(instance_file, output_file, time_limit):

    nb_tasks, nb_resources, capacity, types, resources, duration, \
            nb_successors, successors, nb_tasks_per_resource, \
              task_index_in_resource, types_in_resource, tasks_in_resource, \
                nb_tasks_per_resource, time_horizon \
                 = read_instance(instance_file)
    
    with hexaly.optimizer.HexalyOptimizer() as optimizer:
        #
        # Declare the optimization model
        #
        model = optimizer.model

        # For each resource, the contents of each batch of tasks performed
        batch_content = [[model.set(nb_tasks_per_resource[r]) 
                          for b in range(nb_tasks_per_resource[r])]
                            for r in range(nb_resources)]
        
        # Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        batch_content_arrays = [model.array(batch_content[r]) for r in range(nb_resources)]

        # All tasks are assigned to a batch
        for r in range(nb_resources):
            model.constraint(model.partition(batch_content_arrays[r]))

        # Each batch must consist of tasks with the same type
        types_in_resource_array = model.array(types_in_resource)
        for r in range(nb_resources):
            resource_type_lambda = model.lambda_function(lambda i: types_in_resource_array[r][i])
            for batch in batch_content[r]:
                model.constraint(model.count( model.distinct( batch, resource_type_lambda ) ) <= 1)

        # Each batch cannot exceed the maximum capacity of the resource
        for r in range(nb_resources):
            for batch in batch_content[r]:
                model.constraint(model.count(batch) <= capacity[r])

        # Interval decisions: time range of each batch of tasks
        batch_interval = [[model.interval(0, time_horizon) 
                           for _ in range(nb_tasks_per_resource[r])]
                            for r in range(nb_resources)]
        batch_interval_arrays = [model.array(batch_interval[r]) for r in range(nb_resources)]
    
        # Non-overlap of batch intervals on the same resource
        for r in range(nb_resources):
            for b in range(1, nb_tasks_per_resource[r]):
                model.constraint(batch_interval[r][b-1] < batch_interval[r][b])
        
        # Interval decisions: time range of each task
        task_interval = [None for _ in range(nb_tasks)]
        for t in range(nb_tasks):
            # Retrieve the batch index and resource for this task
            r = resources[t]
            b = model.find( batch_content_arrays[r], task_index_in_resource[t] )
            # Task interval interval associated with task t
            task_interval[t] = batch_interval_arrays[r][b]

        # Task durations
        for t in range(nb_tasks):
            model.constraint(model.length(task_interval[t]) == duration[t])

        # Precedence constraints between tasks
        for t in range(nb_tasks):
            for s in successors[t]:
                model.constraint( task_interval[t] < task_interval[s])

        # Makespan: end of the last task
        makespan = model.max([model.end(model.at(batch_interval_arrays[r], i))
                                for i in range(nb_tasks_per_resource[r]) 
                                    for r in range(nb_resources)])
        model.minimize(makespan)
        model.close()

        # Parametrize the optimizer
        optimizer.param.time_limit = time_limit
        optimizer.solve()

        #
        # Write the solution in a file with the following format:
        #  - makespan
        #  - machine number
        #  - preceeding lines are the ordered intervals of the tasks 
        #    for the corresponding machine number
        if output_file is not None:
            with open(output_file, 'w') as f:
                f.write(str(makespan.value) + "\n")
                for r in range(nb_resources):
                    f.write(str(r) + "\n")
                    for b in range(nb_tasks_per_resource[r]):
                        t = tasks_in_resource[r][b]
                        line = str(t) + " " + str(task_interval[t].value.start()) + " " + str(task_interval[t].value.end())
                        f.write(line + "\n")
            print("Solution written in file ", output_file)


if __name__=="__main__":
    if len(sys.argv) < 2:
        print(
            "Usage: python batch_scheduling.py instance_file [output_file] [time_limit]")
        sys.exit(1)
    
    instance_file = sys.argv[1]
    output_file = sys.argv[2] if len(sys.argv) >= 3 else None
    time_limit = int(sys.argv[3]) if len(sys.argv) >= 4 else 60

    main(instance_file, output_file, time_limit)
Compilation / Execution (Windows)
cl /EHsc batch_scheduling.cpp -I%HX_HOME%\include /link %HX_HOME%\bin\hexaly130.lib
batch_scheduling instances\tai2020_5.txt
Compilation / Execution (Linux)
g++ batch_scheduling.cpp -I/opt/hexaly_13_0/include -lhexaly130 -lpthread -o batch_scheduling
./batch_scheduling instances/tai2020_5.txt
#include "optimizer/hexalyoptimizer.h"
#include <algorithm>
#include <fstream>
#include <iostream>
#include <limits>
#include <numeric>
#include <vector>

using namespace hexaly;
using namespace std;

class BatchScheduling {
private:
    // Hexaly Optimizer
    HexalyOptimizer optimizer;
    // Decision variables: collection of sets (representing batches)
    vector<vector<HxExpression>> batchContent;
    // Decision variables: intervals of batches on a resource
    vector<vector<HxExpression>> batchInterval;
    // Decision variables: interval of a task
    vector<HxExpression> taskInterval;
    // Objective = minimize the makespan: end of the last task of the last job
    HxExpression makespan;

public:
    BatchScheduling() {}

    // Number of tasks
    int nbTasks;
    // Number of resources
    int nbResources;
    // Capacity of each resource
    vector<int> capacity;

    // Number of tasks assigned to each resource
    vector<int> nbTasksPerResource;
    // Types of tasks assigned to each resource
    vector<vector<int>> typesInResources;
    // Tasks assigned to each resource
    vector<vector<int>> tasksInResource;
    // Index of task i on resource[i]
    vector<int> taskIndexInResource;

    // Type of task i
    vector<int> type;
    // Resource required for task i
    vector<int> resource;
    // Duration of task i
    vector<int> duration;
    // Number of tasks that must succeed task i
    vector<int> nbSuccessors;
    // Task ids that must succeed task i
    vector<vector<int>> successors;

    // Longest possible time horizon
    int timeHorizon{0};

    // Read input file
    void readInstance(const string& fileName) {
        ifstream infile;
        infile.exceptions(ifstream::failbit | ifstream::badbit);
        infile.open(fileName.c_str());

        // first line has number of tasks, number of resources
        infile >> nbTasks;
        infile >> nbResources;

        // second line has the capacity of each resource
        capacity.resize(nbResources);
        for (int i = 0; i < nbResources; i++) {
            infile >> capacity[i];
        }
        // initalize
        nbTasksPerResource.resize(nbResources);
        for (int j = 0; j < nbResources; j++) {
            typesInResources.push_back(vector<int>());
            tasksInResource.push_back(vector<int>());
            nbTasksPerResource[j] = 0;
        }
        type.resize(nbTasks);
        resource.resize(nbTasks);
        duration.resize(nbTasks);
        nbSuccessors.resize(nbTasks);
        taskIndexInResource.resize(nbTasks);

        // Task information: [type, machine, duration, nbSuccessors, [successors]]
        for (int i = 0; i < nbTasks; i++) {
            infile >> type[i];
            infile >> resource[i];
            infile >> duration[i];
            infile >> nbSuccessors[i];
            // collect which tasks that must succeed task i
            successors.push_back(vector<int>());
            for (int j = 0; j < nbSuccessors[i]; j++) {
                successors[i].push_back(j);
                infile >> successors[i][j];
            }

            // Index of task i on resource[i]
            taskIndexInResource[i] = nbTasksPerResource[resource[i]];
            // Map from name of task i on resource[i] to task i type
            typesInResources[resource[i]].push_back(type[i]);
            // Map from name of task i on resource[i] to task i
            tasksInResource[resource[i]].push_back(i);
            // Incremenet number of tasks required by this resource
            nbTasksPerResource[resource[i]] += 1;

            // Add task time to the overall trivial time horizon
            timeHorizon += duration[i];
        }
        
        infile.close();
    }

    void solve(int timeLimit) {
        // Declare the optimization model
        HxModel model = optimizer.getModel();

        // For each resource, the contents of each batch of tasks performed
        batchContent.resize(nbResources);
        for (unsigned int r = 0; r < nbResources; r++) {
            batchContent[r].resize(nbTasksPerResource[r]);
            for (unsigned int b = 0; b < nbTasksPerResource[r]; b++) {
                batchContent[r][b] = model.setVar(nbTasksPerResource[r]);
            }
        }

        // Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        vector<HxExpression> batchContentArray = vector<HxExpression>(nbResources);
        for (unsigned int r = 0; r < nbResources; r++) {
            batchContentArray[r] = model.array();
            for (unsigned int b = 0; b < nbTasksPerResource[r]; b++) {
                batchContentArray[r].addOperand(batchContent[r][b]);
            }
        }

        // All tasks are assigned to a batch
        for (unsigned int r = 0; r < nbResources; r++) {
            model.constraint(model.partition(batchContentArray[r]));
        }

        // Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        vector<HxExpression> typesInResourcesArray = vector<HxExpression>(nbResources);
        for (unsigned int r = 0; r < nbResources; r++) {
            typesInResourcesArray[r] = model.array();
            for (unsigned int i = 0; i < nbTasksPerResource[r]; i++) {
                typesInResourcesArray[r].addOperand(typesInResources[r][i]);
            }
        }

        // Each batch must consist of tasks with the same type
        for (unsigned int r = 0; r < nbResources; r++) {
            HxExpression resourceTypeLambda = model.createLambdaFunction([&](HxExpression i) {
                return typesInResourcesArray[r][i];
            });
            for (unsigned int b = 0; b < nbTasksPerResource[r]; b++) {
                HxExpression batch = batchContent[r][b];
                model.constraint(model.count(model.distinct(batch, resourceTypeLambda)) <= 1);
            }
        }

        // Each batch cannot exceed the maximum capacity of the resource
        for (unsigned int r = 0; r < nbResources; r++) {
            for (unsigned int b = 0; b < nbTasksPerResource[r]; b++) {
                model.constraint(model.count(batchContent[r][b]) <= capacity[r]);
            }           
        }

        // Interval decisions: time range of each batch of tasks
        batchInterval.resize(nbResources);
        for (unsigned int r = 0; r < nbResources; r++) {
            batchInterval[r].resize(nbTasksPerResource[r]);
            for (unsigned int b = 0; b < nbTasksPerResource[r]; b++) {
                batchInterval[r][b] = model.intervalVar(0, timeHorizon);
            }
        }

        // Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        vector<HxExpression> batchIntervalArray = vector<HxExpression>(nbResources);
        for (unsigned int r = 0; r < nbResources; r++) {
            batchIntervalArray[r] = model.array();
            for (unsigned int b = 0; b < nbTasksPerResource[r]; b++) {
                batchIntervalArray[r].addOperand(batchInterval[r][b]);
            }
        }

        // Non-overlap of batch intervals on the same resource
        for (unsigned int r = 0; r < nbResources; r++) {
            for (unsigned int b = 1; b < nbTasksPerResource[r]; b++) {
                model.constraint(batchInterval[r][b-1] < batchInterval[r][b]);
            }
        }

        // Interval decisions: time range of each task
        taskInterval.resize(nbTasks);
        for (unsigned int t = 0; t < nbTasks; t++) {
            // Retrieve the batch index and resource for this task
            int r = resource[t];
            HxExpression b = model.find(batchContentArray[r], taskIndexInResource[t]);
            // Task interval associated with task t
            taskInterval[t] = batchIntervalArray[r][b];
        }

        // Task durations
        for (unsigned int t = 0; t < nbTasks; t++) {
            model.constraint(model.length(taskInterval[t]) == duration[t]);
        }

        // Precedence constraints between tasks
        for (unsigned int t = 0; t < nbTasks; t++) {
            for (unsigned int s = 0; s < nbSuccessors[t]; s++) {
                model.constraint(taskInterval[t] < taskInterval[successors[t][s]]);
            }
        }

        // Makespan: end of the last task
        makespan = model.max();
        for (unsigned int t = 0; t < nbTasks; t++) {
            makespan.addOperand(model.end(taskInterval[t]));
        }
        model.minimize(makespan);
        model.close();

        // Parameterize the optimizer
        optimizer.getParam().setTimeLimit(timeLimit);
        optimizer.solve();
    }

    /*
     Write the solution in a file with the following format:
      - makespan
      - machine number
      - preceeding lines are the ordered intervals of the tasks 
        for the corresponding machine number
    */
    void writeSolution(const string& fileName) {
        
        ofstream outfile;
        outfile.exceptions(ofstream::failbit | ofstream::badbit);
        outfile.open(fileName.c_str());
        
        outfile << makespan.getValue() << endl;

        for (unsigned int r = 0; r < nbResources; r++) {
            outfile << r << endl;
            for (unsigned int b = 0; b < nbTasksPerResource[r]; b++) {
                int t = tasksInResource[r][b];
                int start = taskInterval[t].getIntervalValue().getStart();
                int end = taskInterval[t].getIntervalValue().getEnd();
                outfile << t << " " << start << " " << end << endl;
            }
        }
        cout << "Solution written in file " << fileName << endl;
        outfile.close();
    }
    
};

int main(int argc, char** argv) {
    if (argc < 2) {
        cout << "Usage: batch_scheduling instanceFile [outputFile] [timeLimit]" << endl;
        exit(1);
    }

    const char* instanceFile = argv[1];
    const char* outputFile = argc > 2 ? argv[2] : NULL;
    const char* strTimeLimit = argc > 3 ? argv[3] : "60";

    BatchScheduling model;
    try {
        model.readInstance(instanceFile);
        const int timeLimit = atoi(strTimeLimit);
        model.solve(timeLimit);
        if (outputFile != NULL)
            model.writeSolution(outputFile);
        return 0;
    } catch (const exception& e) {
        cerr << "An error occurred: " << e.what() << endl;
        return 1;
    }
}
Compilation / Execution (Windows)
copy %HX_HOME%\bin\Hexaly.NET.dll .
csc BatchScheduling.cs /reference:Hexaly.NET.dll
BatchScheduling instances\tai2020_5.txt
using System;
using System.IO;
using Hexaly.Optimizer;

public class BatchScheduling : IDisposable
{
    // Number of tasks
    private int nbTasks;
    // Number of resources
    private int nbResources;
    // Capacity of each resource
    private int[] capacity;

    // Number of tasks assigned to each resource
    private int[] nbTasksPerResource;
    // Types of tasks assigned to each resource
    private int[][] typesInResources;
    // Tasks assigned to each resource
    private int[][] tasksInResource;
    // Index of task i on resource[i]
    private int[] taskIndexInResource;

    // Type of task i
    private int[] type;
    // Resource required for task i
    private int[] resource;
    // Duration of task i
    private int[] duration;
    // Number of tasks that must succeed task i
    private int[] nbSuccessors;
    // Task names that must succeed task i
    private int[][] successors;

    // Longest possible time horizon
    private int timeHorizon;

    // Hexaly Optimizer
    private HexalyOptimizer optimizer;
    // Decision variables: collection of sets (representing batches)
    private HxExpression[][] batchContent;
    // Decision variables: intervals of batches on a resource
    private HxExpression[][] batchInterval;
    // Decision variables: interval of a task
    private HxExpression[] taskInterval;
    // Objective = minimize the makespan: end of the last task of the last job
    private HxExpression makespan;

    public BatchScheduling(string fileName)
    {
        optimizer = new HexalyOptimizer();
    }
    public static string[] ReadNextLine(StreamReader input) {
        return input.ReadLine().Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
    }

    public void ReadInstance(string fileName) 
    {
        using (StreamReader input = new StreamReader(fileName))
        {
            string[] splitted = ReadNextLine(input);

            // first line has number of tasks, number of resources
            nbTasks = int.Parse(splitted[0]);
            nbResources = int.Parse(splitted[1]);

            // second line has the capacity of each resource
            capacity = new int[nbResources];
            splitted = ReadNextLine(input);
            for (int r = 0; r < nbResources; ++r) {
                capacity[r] = int.Parse(splitted[r]);
            }
            // initialize
            type = new int[nbTasks];
            resource = new int[nbTasks];
            duration = new int[nbTasks];
            nbSuccessors = new int[nbTasks];
            successors = new int[nbTasks][];
            taskIndexInResource = new int[nbTasks];
            typesInResources = new int[nbResources][];
            tasksInResource = new int[nbResources][];
            nbTasksPerResource = new int[nbResources];
            for (int j = 0; j < nbResources; ++j)
                nbTasksPerResource[j] = 0;

            // Task information: [type, machine, duration, nbSuccessors, [successors]]
            for (int i = 0; i < nbTasks; ++i) 
            {
                splitted = ReadNextLine(input);

                type[i] = int.Parse(splitted[0]);
                resource[i] = int.Parse(splitted[1]);
                duration[i] = int.Parse(splitted[2]);
                nbSuccessors[i] = int.Parse(splitted[3]);

                // collect which tasks that must succeed task i
                successors[i] = new int[nbSuccessors[i]];
                for (int j = 0; j < nbSuccessors[i]; j++) {
                    successors[i][j] = int.Parse(splitted[4+j]);
                }

                // Index of task i on resource[i]
                taskIndexInResource[i] = nbTasksPerResource[resource[i]];
                // Incremenet number of tasks required by this resource
                nbTasksPerResource[resource[i]] += 1;

                // Add task time to the overall trivial time horizon
                timeHorizon += duration[i];
            }

            // Mapping task list to resource task lists
            int[] temp_nbTasksPerResource = new int[nbResources];
            for (int r = 0; r < nbResources; ++r)
            {
                typesInResources[r] = new int[nbTasksPerResource[r]];
                tasksInResource[r] = new int[nbTasksPerResource[r]];
                temp_nbTasksPerResource[r] = 0;
            }
            for (int i = 0; i < nbTasks; ++i) {
                int r = resource[i];
                int index = temp_nbTasksPerResource[r];
                // Map from name of task i on resource[i] to task i type
                typesInResources[r][index] = type[i];
                // Map from name of task i on resource[i] to task i
                tasksInResource[r][index] = i;

                // increment
                temp_nbTasksPerResource[r] += 1;
            }
        }
    }

    public void Dispose()
    {
        if (optimizer != null)
            optimizer.Dispose();
    }

    public void Solve(int timeLimit)
    {
        // Declare the optimization model
        HxModel model = optimizer.GetModel();

        // For each resource, the contents of each batch of tasks performed
        batchContent = new HxExpression[nbResources][];
        for (int r = 0; r < nbResources; ++r)
        {
            batchContent[r] = new HxExpression[nbTasksPerResource[r]];
            for (int b = 0; b < nbTasksPerResource[r]; ++b)
                batchContent[r][b] = model.Set(nbTasksPerResource[r]);
        }

        // Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        HxExpression[] batchContentArray = new HxExpression[nbResources];
        for (int r = 0; r < nbResources; ++r)
            batchContentArray[r] = model.Array(batchContent[r]);

        // All tasks are assigned to a batch
        for (int r = 0; r < nbResources; ++r)
            model.Constraint( model.Partition(batchContentArray[r]) );

        // Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        HxExpression[] typesInResourcesArray = new HxExpression[nbResources];
        for (int r = 0; r < nbResources; ++r) 
            typesInResourcesArray[r] = model.Array(typesInResources[r]);

        // Each batch must consist of tasks with the same type
        for (int r = 0; r < nbResources; ++r) {
            HxExpression resourceTypeLambda = model.LambdaFunction( i =>
            {
                return typesInResourcesArray[r][i];
            });
            for (int b = 0; b < nbTasksPerResource[r]; ++b)
                model.Constraint(model.Count( 
                    model.Distinct(batchContent[r][b], resourceTypeLambda)) <= 1);
        }

        // Each batch cannot exceed the maximum capacity of the resource
        for (int r = 0; r < nbResources; ++r)
        {
            for (int b = 0; b < nbTasksPerResource[r]; ++b)
                model.Constraint(model.Count(batchContent[r][b]) <= capacity[r]);
        }

        // Interval decisions: time range of each batch of tasks
        batchInterval = new HxExpression[nbResources][];
        for (int r = 0; r < nbResources; ++r)
        {
            batchInterval[r] = new HxExpression[nbTasksPerResource[r]];
            for (int b = 0; b < nbTasksPerResource[r]; ++b)
                batchInterval[r][b] = model.Interval(0, timeHorizon);
        }

        // Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        HxExpression[] batchIntervalArray = new HxExpression[nbResources];
        for (int r = 0; r < nbResources; ++r)
            batchIntervalArray[r] = model.Array(batchInterval[r]);

        // Non-overlap of batch intervals on the same resource
        for (int r = 0; r < nbResources; ++r)
        {
            for (int b = 1; b < nbTasksPerResource[r]; ++b)
                model.Constraint(batchInterval[r][b-1] < batchInterval[r][b]);
        }

        // Interval decisions: time range of each task
        taskInterval = new HxExpression[nbTasks];
        for (int t = 0; t < nbTasks; ++t) {
            // Retrieve the batch index and resource for this task
            int r = resource[t];
            HxExpression b = model.Find(batchContentArray[r], taskIndexInResource[t]);
            // Task interval associated with task t
            taskInterval[t] = batchIntervalArray[r][b];
        }

        // Task durations
        for (int t = 0; t < nbTasks; ++t)
            model.Constraint(model.Length(taskInterval[t]) == duration[t]);

        // Precedence constraints between tasks
        for (int t = 0; t < nbTasks; ++t)
        {
            for (int s = 0; s < nbSuccessors[t]; ++s)
                model.Constraint(taskInterval[t] < taskInterval[successors[t][s]]);
        }

        // Makespan: end of the last task
        makespan = model.Max();
        for (int t = 0; t < nbTasks; ++t)
            makespan.AddOperand(model.End(taskInterval[t]));
        model.Minimize(makespan);
        model.Close();

        // Parameterize the optimizer
        optimizer.GetParam().SetTimeLimit(timeLimit);

        optimizer.Solve();
    }

    /* Write the solution in a file with the following format:
     *  - makespan
     *  - machine number
     *  - preceeding lines are the ordered intervals of the tasks 
          for the corresponding machine number */
    public void WriteSolution(string fileName)
    {
        using (StreamWriter output = new StreamWriter(fileName))
        {
            Console.WriteLine("Solution written in file " + fileName);
            output.WriteLine(makespan.GetValue());
            for (int r = 0; r < nbResources; ++r)
            {
                output.Write(r);
                output.WriteLine();
                for (int b = 0; b < nbTasksPerResource[r]; ++b)
                {
                    int t = tasksInResource[r][b];
                    long start = taskInterval[t].GetIntervalValue().Start();
                    long end = taskInterval[t].GetIntervalValue().End();
                    output.Write(t + " " + start + " " + end);
                    output.WriteLine();
                }
            }
        }
    }

    public static void Main(string[] args)
    {
        if (args.Length < 1)
        {
            Console.WriteLine("Usage: batching_scheduling.cs instanceFile [outputFile] [timeLimit]");
            System.Environment.Exit(1);
        }

        string instanceFile = args[0];
        string outputFile = args.Length > 1 ? args[1] : null;
        string strTimeLimit = args.Length > 2 ? args[2] : "60";

        using (BatchScheduling model = new BatchScheduling(instanceFile))
        {
            model.ReadInstance(instanceFile);
            model.Solve(int.Parse(strTimeLimit));
            if (outputFile != null)
                model.WriteSolution(outputFile);
        }
    }
}
Compilation / Execution (Windows)
javac BatchScheduling.java -cp %HX_HOME%\bin\hexaly.jar
java -cp %HX_HOME%\bin\hexaly.jar;. BatchScheduling instances\tai2020_5.txt
Compilation / Execution (Linux)
javac BatchScheduling.java -cp /opt/hexaly_13_0/bin/hexaly.jar
java -cp /opt/hexaly_13_0/bin/hexaly.jar:. BatchScheduling instances/tai2020_5.txt
import java.util.*;
import java.io.*;
import com.hexaly.optimizer.*;

public class BatchScheduling {
    // Number of tasks
    private int nbTasks;
    // Number of resources
    private int nbResources;
    // Capacity of each resource
    private int[] capacity;

    // Number of tasks assigned to each resource
    private int[] nbTasksPerResource;
    // Types of tasks assigned to each resource
    private int[][] typesInResources;
    // Tasks assigned to each resource
    private int[][] tasksInResource;
    // Index of task i on resource[i]
    private int[] taskIndexInResource;

    // Type of task i
    private int[] type;
    // Resource required for task i
    private int[] resource;
    // Duration of task i
    private int[] duration;
    // Number of tasks that must succeed task i
    private int[] nbSuccessors;
    // Task names that must succeed task i
    private int[][] successors;

    // Longest possible time horizon
    private int timeHorizon;

    // Hexaly Optimizer
    final HexalyOptimizer optimizer;
    // Decision variables: collection of sets (representing batches)
    private HxExpression[][] batchContent;
    // Decision variables: intervals of batches on a resource
    private HxExpression[][] batchInterval;
    // Decision variables: interval of a task
    private HxExpression[] taskInterval;
    // Objective = minimize the makespan: end of the last task of the last job
    private HxExpression makespan;

    public BatchScheduling(HexalyOptimizer optimizer, String fileName) throws IOException
    {
        this.optimizer = new HexalyOptimizer();
    }

    private void readInstance(String fileName) throws IOException {
        try (Scanner input = new Scanner(new File(fileName))) {
            
            // first line has number of tasks, number of resources
            nbTasks = input.nextInt();
            nbResources = input.nextInt();

            // second line has the capacity of each resource
            capacity = new int[nbResources];
            for (int r = 0; r < nbResources; ++r) {
                capacity[r] = input.nextInt();
            }
            
            // initialize
            type = new int[nbTasks];
            resource = new int[nbTasks];
            duration = new int[nbTasks];
            nbSuccessors = new int[nbTasks];
            successors = new int[nbTasks][];
            taskIndexInResource = new int[nbTasks];
            typesInResources = new int[nbResources][];
            tasksInResource = new int[nbResources][];
            nbTasksPerResource = new int[nbResources];
            for (int j = 0; j < nbResources; ++j)
                nbTasksPerResource[j] = 0;

            // Task information: [type, machine, duration, nbSuccessors, [successors]]
            for (int i = 0; i < nbTasks; i++) {

                type[i] = input.nextInt();
                resource[i] = input.nextInt();
                duration[i] = input.nextInt();
                nbSuccessors[i] = input.nextInt();

                // collect which tasks that must succeed task i
                successors[i] = new int[nbSuccessors[i]];
                for (int j = 0; j < nbSuccessors[i]; j++) {
                    successors[i][j] = input.nextInt();
                }

                // Index of task i on resource[i]
                taskIndexInResource[i] = nbTasksPerResource[resource[i]];
                // Incremenet number of tasks required by this resource
                nbTasksPerResource[resource[i]] += 1;

                // Add task time to the overall trivial time horizon
                timeHorizon += duration[i];
            }

            // Mapping task list to resource task lists
            int[] temp_nbTasksPerResource = new int[nbResources];
            for (int r = 0; r < nbResources; ++r)
            {
                typesInResources[r] = new int[nbTasksPerResource[r]];
                tasksInResource[r] = new int[nbTasksPerResource[r]];
                temp_nbTasksPerResource[r] = 0;
            }
            for (int i = 0; i < nbTasks; ++i) {
                int r = resource[i];
                int index = temp_nbTasksPerResource[r];
                // Map from name of task i on resource[i] to task i type
                typesInResources[r][index] = type[i];
                // Map from name of task i on resource[i] to task i
                tasksInResource[r][index] = i;

                // increment
                temp_nbTasksPerResource[r] += 1;
            }
        }
    }

    public void solve(int timeLimit) {
        // Declare the optimization model
        HxModel model = optimizer.getModel();

        // For each resource, the contents of each batch of tasks performed
        batchContent = new HxExpression[nbResources][];
        for (int r = 0; r < nbResources; ++r)
        {
            batchContent[r] = new HxExpression[nbTasksPerResource[r]];
            for (int b = 0; b < nbTasksPerResource[r]; ++b)
                batchContent[r][b] = model.setVar(nbTasksPerResource[r]);
        }

        // Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        HxExpression[] batchContentArray = new HxExpression[nbResources];
        for (int r = 0; r < nbResources; ++r)
            batchContentArray[r] = model.array(batchContent[r]);

        // All tasks are assigned to a batch
        for (int r = 0; r < nbResources; ++r)
            model.constraint(model.partition(batchContentArray[r]));
        
        // Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        HxExpression[] typesInResourcesArray = new HxExpression[nbResources];
        for (int r = 0; r < nbResources; ++r) 
            typesInResourcesArray[r] = model.array(typesInResources[r]);

        // Each batch must consist of tasks with the same type
        for (int r = 0; r < nbResources; ++r) {
            final int rL = r;
            HxExpression resourceTypeLambda = model.lambdaFunction( i -> {
                return model.at(typesInResourcesArray[rL], i);
            });
            for (int b = 0; b < nbTasksPerResource[r]; ++b)
                model.constraint(model.leq(model.count(model.distinct(batchContent[r][b], resourceTypeLambda)), 1));
        }

        // Each batch cannot exceed the maximum capacity of the resource
        for (int r = 0; r < nbResources; ++r)
        {
            for (int b = 0; b < nbTasksPerResource[r]; ++b)
                model.constraint(model.leq(model.count(batchContent[r][b]), capacity[r]));
        }

        // Interval decisions: time range of each batch of tasks
        batchInterval = new HxExpression[nbResources][];
        for (int r = 0; r < nbResources; ++r)
        {
            batchInterval[r] = new HxExpression[nbTasksPerResource[r]];
            for (int b = 0; b < nbTasksPerResource[r]; ++b)
                batchInterval[r][b] = model.intervalVar(0, timeHorizon);
        }

        // Create HexalyOptimizer arrays in order to be able to access them with "at" operators
        HxExpression[] batchIntervalArray = new HxExpression[nbResources];
        for (int r = 0; r < nbResources; ++r)
            batchIntervalArray[r] = model.array(batchInterval[r]);

        // Non-overlap of batch intervals on the same resource
        for (int r = 0; r < nbResources; ++r)
        {
            for (int b = 1; b < nbTasksPerResource[r]; ++b)
                model.constraint(model.lt(batchInterval[r][b-1], batchInterval[r][b]));
        }

        // Interval decisions: time range of each task
        taskInterval = new HxExpression[nbTasks];
        for (int t = 0; t < nbTasks; ++t) {
            // Retrieve the batch index and resource for this task
            int r = resource[t];
            HxExpression b = model.find(batchContentArray[r], taskIndexInResource[t]);
            // Task interval associated with task t
            taskInterval[t] = model.at(batchIntervalArray[r], b);
        }

        // Task durations
        for (int t = 0; t < nbTasks; ++t)
            model.constraint(model.eq(model.length(taskInterval[t]), duration[t]));

        // Precedence constraints between tasks
        for (int t = 0; t < nbTasks; ++t)
        {
            for (int s = 0; s < nbSuccessors[t]; ++s)
                model.constraint(model.lt(taskInterval[t], taskInterval[successors[t][s]]));
        }
        
        // Makespan: end of the last task
        makespan = model.max();
        for (int t = 0; t < nbTasks; ++t)
            makespan.addOperand(model.end(taskInterval[t]));
        model.minimize(makespan);
        model.close();

        // Parameterize the optimizer
        optimizer.getParam().setTimeLimit(timeLimit);

        optimizer.solve();
    }

    /*
     * Write the solution in a file with the following format:
     * - makespan
     * - machine number
     * - preceeding lines are the ordered intervals of the tasks 
          for the corresponding machine number
     */
    public void writeSolution(String fileName) throws IOException {
        try (PrintWriter output = new PrintWriter(fileName)) {
            System.out.println("Solution written in file " + fileName);

            output.write(makespan.getValue() + "\n");
            for (int r = 0; r < nbResources; ++r) {
                output.write(r + "\n");
                for (int b = 0; b < nbTasksPerResource[r]; ++b) {
                    int t = tasksInResource[r][b];
                    long start = taskInterval[t].getIntervalValue().getStart();
                    long end = taskInterval[t].getIntervalValue().getEnd();
                    output.write(t + " " + start + " " + end + "\n");
                }
            }
        }
    }

    public static void main(String[] args) {
        if (args.length < 1) {
            System.out.println("Usage: java BatchScheduling instanceFile [outputFile] [timeLimit]");
            System.exit(1);
        }

        String instanceFile = args[0];
        String outputFile = args.length > 1 ? args[1] : null;
        String strTimeLimit = args.length > 2 ? args[2] : "60";

        try (HexalyOptimizer optimizer = new HexalyOptimizer()) {
            BatchScheduling model = new BatchScheduling(optimizer, instanceFile);
            model.readInstance(instanceFile);
            model.solve(Integer.parseInt(strTimeLimit));
            if (outputFile != null) {
                model.writeSolution(outputFile);
            }
        } catch (Exception ex) {
            System.err.println(ex);
            ex.printStackTrace();
            System.exit(1);
        }
    }
}