Objective Q not PSD (diagonal adjustment would be required)

[Heinu Schütte]

Hello,

I am trying to solve my first model using Gurobi in Python.

It is a very simple minimisation problem and everything works as expected, until I add this constraint:
c4 = m.addConstr(l, GRB.EQUAL, 16.0373*w + 31.9586)

I am trying to set the load hauled by a locomotive equal to 16.0373*(the number of wagons) + 31.9586.

When running the model, I get:

gurobipy.GurobiError: Objective Q not PSD (diagonal adjustment of 1.8e+03 would be required)

I have read some other questions regarding the "Objective Q not PSD" error, but I don't understand what it means, nor how to fix it.

Any help will be greatly appreciated.

Regards,

Heinu

[Daniel Espinoza]

Hi Heinu,

Assuming that `w` and `I` are variables, did you try 

c4 = m.addConstr(16.0373*w + 31.9586, GRB.EQUAL, I)

instead?

or even simpler:

c4 = m.addConstr(I == 16.0373*w + 31.9586)

Let me know if they work

Best

Daniel

[Daniel Espinoza]

Still,

It seems strange to me that you are getting that behavior, could you share a more complete example of the problem?

Best

Daniel

[Heinu Schütte]

Hello Daniel,

Thank you for the response. I tried

c4 = m.addConstr(I == 16.0373*w + 31.9586)

and that seems to work.

However, now I get "gurobipy.GurobiError: Unable to retrieve attribute 'X' ". Which goes away when I take c4 out - but I get a useless calculation result.

Below is the full code:

from gurobipy import *

# Set calculation-loop lists

days = range(1,366)

speedlist = range(5,115,5)

# Repeat entire optimisation calculation for every increment of train operating

# speed = {5;10;15;...;110}

for i in speedlist:

    v = i

    print

    print "--- --- --- --- ---"

    print

    print "For train operating speed set to", v, "km/h:"

    print

    print "--- --- --- --- ---"

    print

    

    # Set calculation values for each iteration of operating speed:

    d = 117.70

    t = d/v

    

    # Create model

    m = Model('Swazilink Optimisation')

    

    # Define variables

    n = m.addVar(vtype=GRB.INTEGER, name='Number of trains per day')

    w = m.addVar(vtype=GRB.INTEGER, name='Number of wagons per train')

    c = m.addVar(vtype=GRB.INTEGER, name='Number of containers per train')

    l = m.addVar(vtype=GRB.CONTINUOUS, name='Load hauled per train')

    

    # Update Gurobi variables

    m.update()

    

    # Define objective function

    obj = quicksum(QuadExpr(n*l*v)

        for i in days)

    yearly = quicksum(n*w

        for j in days)

    

    # Maximise minimise objective function        

    m.setObjective(obj, GRB.MINIMIZE)

    

    # Set constraints

    c1 = m.addConstr(w, GRB.LESS_EQUAL, 50)

    c2 = m.addConstr(c, GRB.EQUAL, 2*w)

    c3 = m.addConstr(l, GRB.EQUAL, w)

    c4 = m.addConstr(l == 16.0373*w + 31.9586)

    c5 = m.addConstr(n*w, GRB.GREATER_EQUAL, 1305)

    c6 = m.addConstr(yearly, GRB.GREATER_EQUAL, 476463)

    c7 = m.addConstr(w, GRB.GREATER_EQUAL, 0)

    c8 = m.addConstr(c, GRB.GREATER_EQUAL, 0)

    

    # Run model optimization

    m.optimize()

    # Print attributes

    m.printAttr('obj')

    m.printAttr('X')

Regards,

Heinu

[Daniel Espinoza]

Dear Heinu,

It seems that your model is infeasible (even in the first instance where speed=5); thus when you query X you get an error. (see R2 and R3 and R4)

Best

Daniel

[Heinu Schütte]

Dear Daniel,

Oh dear. Thank you very much for your assistance.

Best regards

Heinu