Originally posted by: SystemAdmin

[nadeera said:]

hi!

I'm very new to Cplex and i have not recieved any training on Cplex..Currently i'm trying to apply the column generation technique to a set partitioning problem. The problem is based on SOMBAT et al paper where we're partitioning the set of items into disjoint groups and each group is assigned to a vehicle. No item can be assigned to more than one group (orders cannot be spplit across multiple vehicles) but a supplier can be visited by multiple vehicles (each supplier produces one or more non-identical items).

Since i'm not that good in programming, i'm using the cutstock example (cutstock_int_main.mod) provided, and alter it to fit my problem. My problem would be:

1. [color=blue]masterData.Alpha.add(masterData.Alpha.truck,1,subOpl.z.solutionValue);[/color]
    this command is taken straight from the cutstock problem but i changed the variable name accordingly. It is to prepare for the next iteration. When i   
    try to run it an error occur : [color=red]Scripting runtime error:Wrong component type at position 0, expecting int found float, "undefined".[/color]
    i know it is straight forward but even if i change all my var to type int the same problem still occur. So, is it the solving engine or what??

2. [color=blue]for(var i in masterOpl.Items) {
      subData.Duals[i] = masterOpl.ctFill[i].dual; }[/color] 
  I believe this command is to get the dual of the primary problem. Cutstock has only one constraint but mine has 2 constraints. My question is, can we actually named 2 constraints under one label or we only have to use the same command twice with different ctlabel?

3. Another major difference for my problem is that my decision variable is binary (hence the use of boolean for my dvar). So im wondering is it possible for me to use the coding for cutstock n apply it to my inventory problem??

Last but not least, i would appreciate it very much if anyone can show me sample coding for column generation technique (esp those using boolean dvar) on VRP..thank you!!
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[alain.chabrier said:]

Hi,

some answers :

1) the structure masterData.Alpha does not correspond to the values you pass to add a tuple.
the first element expected is an int and you pass a float.

2) yes, one such call per constraint.

Alain
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[nadeera said:]

thank you alain! It's enough to give me some idea..I appreciate it
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: Mathsg

Hi,

I am solving networking planing problem by column generation approache in which i have 6 node topology. Source and destination are connected through k shortest path.

i have decompose the problem into master and sub problem. i am going to minimize the cost of flow. In the master problem i have cosider 2 k shortest path for each flow between source and destination which give me current solution for the flows and sub problem can generate new path which can imporve our objective i.e., minimize the cost. path is made up of connecting different link and combine them i.e., hope count. Now i am confused that how to write the next iteration for updating the master problem as i have get reduced cost from sub problem which means there a path i.e., links that can improve our master problem objective. i have written it but it give me an error. can you help me ? it would be highly appricated and i am looking forward to hearing from you. thank u

#DecisionOptimization
#OPLusingCPLEXOptimizer