Issues with single resource flowtime optimizer

[dreinon]

The first issue I have ran into by trying out multiple single resource flowtime objectives is that the value of all of them is the same. For instance, I get the following values: ->Objective values: ->FlowtimeSingleResource(min objective): 9 ->FlowtimeSingleResource(min objective): 9 for this result: where I'm optimizing flowtime between time intervals (11, 20) and (21, 34), so total in-interval flowtime is 0. I implemented a way to calculate this total value that doesn't rely on indicator values though.

[dreinon]

I don't get any found optimum, I just get

Found better value: -8552 elapsed time:179.280s
Checking better value <-8552

I'll leave it for 1200 sec (20 min) and see if it gets any better.

[dreinon]

I could also add the constraint for debugging but in real life I would need it to be automatic.

[dreinon]

Output is:

/xxxx/ProcessScheduler/processscheduler/solver.py:434: UserWarning: time may exceed max time. Stopping iteration.
  warnings.warn('time may exceed max time. Stopping iteration.')
        total number of iterations: 139
        value: -8830
        XXXXXXXXXXXXX satisfiability checked in 953.49s

with occupation of 97

[tpaviot]

Maybe the solver is lost into some kind of local optimum. You can try other logics to see if it is better ("QF_IDL", "QF_UFIDL" or others)

[dreinon]

Okay, thanks!

[dreinon]

I have tried solving first for utilization objective only and then, solving again, but with a constraint that utilization indicator has to be equal to the previously found value, and with single resource flowtime objectives, and I get a solution with 100% occupation and pretty flowtime optimized :)

[dreinon]

Maybe an option for multiobjective incremental optimizer would be to solve for objectives in creation order (equivalent to 'lexicon' option).

[tpaviot]

I agree, that's something I think about.

[dreinon]

Closing this issue and opening one for this new feature #73.