Originally posted by: trxw

Hi Alex

I need your expertice once more... 

I have model a rostering problem with binaries integer variables the following way.

- Shift is divided in slots (1 to 96) (each slot is 5 minutes but thats not relevant for CP).

- With up to 24 people I need to fill in up to 18 positions. Not all the positions are opened all the time and I don´t always have 30 people.

- Rules are that a person cannot work more than 12 slots in a 150 slot sliding window.

- Working time should be distributed evenly among the workers

- The less number of breaks the better, meaning that it is better to work 120 / rest 30, than work 40/ rest 40/ work 40 / rest 30 for example

My model works quite alright with up to 4 positions and 7 workers finding a solution in less than a minute. If I tried to go for example to find a solution for 6 positions and 9 people, it's unable to find one in 5 minutes.

Do you think my aproach of using integer variables and slots is correct or should I go for interval time variables instead for better performance?

I have tried using something called tunning tool in order to find better parameter settings for my model but I haven´t got too much success yet...

I have shared with you the files

Thanks

Hi,

do not hesitate to post your model and data so that other users could have a look.

Do you use MIP or CP ?

regards

Originally posted by: trxw

Hi,

I use CP. Find attached files. Thanks for your inputs

Hi,

even if I change your model into MIP as attached file I do not get any solution after 10 minutes

So I wonder whether your model is feasible.

regards

Originally posted by: trxw

It is feasible. See attached. I got the solution manually...

The problem is that it's quite hard to see if there's a problem in the model. Is there any way to debug the model?