Creating model takes too much time

Ongoing

Angyang Chen

• October 08, 2020 04:25
• Edited

 

Hi All

I have a problem in a big scale network modeling.

I write the sets as

S = [s for s in range(1, 7)]
S_ = [s_ for s_ in range(1, 7)]
T = [t for t in range(0, 145)]
T_ = [t_ for t_ in range(0, 145)]
T1 = [t1 for t1 in range(0, 145)]
T1_ = [t1_ for t1_ in range(0, 145)]

P = [p for p in range(1, 3)]
P_ = [p_ for p_ in range(1, 3)]

K = [k for k in range(0, 17)]
K_ = [k_ for k_ in range(0, 17)]

L = [l for l in range(1, 14)]
L_ = [l_ for l_ in range(1, 14)]

I write the variables as

x = m.addVars(L, P, K, arc_set.keys(), vtype=GRB.BINARY, name="x")

I write the constraints as
for s in S:
    for s_ in S_:
       for t in T:
           if s_ == s + 1:
               m.addConstr(
                   gp.quicksum(x[l, p, kt, s, s_, t1, t1_] for l in L for p in P for kt in KT for t1 in T1 for t1_ in T1_
                   if stop_pattern[kt, s] == 1 and t1_ == t1 + real_runtime[p, kt, s, s_] and (s, s_, t1, t1_) in segment_arc.keys() and t <= t1 < t + ff_headway[s]),
                   GRB.LESS_EQUAL,
                   1,
                   'fafa_headway')

It takes 2000 seconds for Gurobi to creat these constraints and I dont know why. Is there any faster way?

1

• Official comment

  

  Simranjit Kaur

  Gurobi Staff

  • December 06, 2024 00:41

  This post is more than three years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum. Or why not try our AI Gurobot?.
• 

  Sonja Mars

  Gurobi Staff

  • October 08, 2020 07:19

  Hi,

  These are many for loops that you have and a lot of conditions. Unfortunately, your code is not self-contained so I could not run it to do some tests. However, you should some small improvement by changing this 

  for s in S:
      for s_ in S_:
         for t in T:
             if s_ == s + 1:

  into this

  for s in S:
      for s_ in S_:
         if s_ == s + 1:
             for t in T:

  Additionally, the conditions for the quicksum seem to be quite complicated. Are you sure you have the best data structures for what you want to do? Is there any chance you can come up with a model that does not have a variable with 7 indices? How many x variables do you have? This seems to be something about arcs? So is this a dense network, do all arcs between two nodes exist? Can you represent this in a sparse way? 

  Best,

    Sonja

  0

Post is closed for comments.