Employee Scheduling example using the CP-SAT solver in Java gives INFEASIBLE status

[Chatree Srichart]

Hello Everyone,

I wrote a class (attached file) to solve the Employee Scheduling in Java as in this Python example:

https://developers.google.com/optimization/scheduling/employee_scheduling

However, the solver gives INFEASIBLE status.

What I need to do to make it OPTIMAL are:

1. change the upper bound of 
   model.addLinearSum(d_shifts, 0, 1);
   from 1 to 2
2. delete this line
   model.addLinearSum(n_shifts_worked, min_shifts_per_nurse, max_shifts_per_nurse);
   
   

However, the solved values are not the same as in the Python example.

Note: I uses the or-tools library which was compiled from the master yesterday.

Any help would be appreciated.

Thanks.

[Chatree Srichart]

For more information

What I did in the Java class (attached file in the original post) to replace the original Python example:

https://developers.google.com/optimization/scheduling/employee_scheduling

are:

From Python

model.Add(min_shifts_per_nurse <= num_shifts_worked)
model.Add(num_shifts_worked <= max_shifts_per_nurse)

To Java

int min_shifts_per_nurse = num_shifts * num_days; // num_nurses
int max_shifts_per_nurse = min_shifts_per_nurse + 1;IntVar[] n_shifts_worked = new IntVar[num_shifts * num_days];
            for (int s : all_shifts) {
                for (int d : all_days) {
                    int index = s * num_days + d;
                    n_shifts_worked[index] = shifts[n][d][s];
                }
            }
model.addLinearSum(n_shifts_worked, min_shifts_per_nurse, max_shifts_per_nurse);

and

From Python

model.Maximize(
        sum(shift_requests[n][d][s] * shifts[(n, d, s)] for n in all_nurses
            for d in all_days for s in all_shifts))

To Java

IntVar[] intVars = new IntVar[num_shifts * num_days * num_nurses];
        int[] ints = new int[num_shifts * num_days * num_nurses];
        for (int s : all_shifts) {
            for (int d : all_days) {
                for (int n : all_nurses) {
                    int index = (s * num_days + d) * num_nurses + n;
                    intVars[index] = shifts[n][d][s];
                    ints[index] = shift_requests[n][d][s];
                }
            }
        }
        model.maximizeScalProd(intVars, ints);

Are they correct?

Note: I still don't know why my EmployeeSchedulingSat.java class (attached file) does not produce the same solution as in the original Python example https://developers.google.com/optimization/scheduling/employee_scheduling .

เมื่อ วันพฤหัสบดีที่ 11 เมษายน ค.ศ. 2019 20 นาฬิกา 14 นาที 37 วินาที UTC+7, Chatree Srichart เขียนว่า: