Forcing vehicle steps in solving mode

[mbachem]

Hi there,

I try to force some steps to a certain vehicle, in solving mode.

I started with three (pickup/delivery) shipments (0/1), (2/3) and (4/5), and two vehicles, vehicle:0 an vehicle:1
All three shipments can be done by each vehicle alone, shown by problem_1.json and problem_2.json.
When both vehicle are at service, vroom is using vehicle:0 for (0/1), and vehicle:1 for (4/5) and (2/3).

Now I try to pin shipment (4/5) to vehicle:0, without any success, as shown by problem_4.json an solution_4.json
What am I doing wrong?

All my files described above can be reviewed here, a vroom forked 'vehicle_pinning' branch :
d1773c8

Thanks a lot for helping!

cheers,
Martin

[mbachem]

Hi There,
I think I have found a pretty capable way of vehicle pinning: when each vehicle brings it's own ID as skill, once a shipment needs a vehicle pinning, I force the desired shipment-vehicle-pinning by requiring that vehicle-skill-ID as shipment skill.
Plus setting a high shipment priority for already pinned shipments, I expect new shipments without vehicle-pinning and low priority, could be easily unassigned when running into conflicts.
Cheers,
Martin

[fbaierl]

Hi mbachem,

this topic has been discussed before already, see e.g.:

#800
#753

Maybe you can find a few helpful hints in those threads.

[mbachem-rvk]

Both topics are indeed super helpfull, thanks a lot for pointing them out. Good to see I'm on the right track by setting vehicle specific skills to render shipments as immutable.

[mbachem-rvk]

Hi @fbaierl ,
I have found some strange unexpected results along the way...
( comitted as https://github.com/mbachem-rvk/vroom/tree/testing/unexpected_behaviour/ ... )

• problem1 is solved as solution1, 0 unassingned, where shipment.2493057 is serviced by vehicle 502
• alas, when I set shipment.2493097 to require skill [502], which is unique for vehicle 502, the shipment renders out unassigned
  • shown in problem2
  • https://github.com/mbachem-rvk/vroom/tree/testing/unexpected_behaviour/v1.13

Then, there is an another strange matter:

• in problem3, I get an unresolved shipment, while I have a lot of completely unoccupied vehicles...
• setting shipment.2492940 to require skill [1201] (in problem3.json), for I know vehicle 1201 with skill 1201 is completely free, will render 0 unassigned again
• this is getting better if each vehicle gets a 'filler' job assigned without load, to tickle the path...

[mbachem-rvk]

... I moved away from vehicle pinning by using vehicle unique skills.
Would ne great to do so, but this but comes with too many strange effects...

As I go along the day, every new incoming shipment-request can either be accepted as servicable, or has to be rejected for good.
To do so, I have a pool of still mutable shipments, where I can accept changes in vehicle allocation, and collect those in seperate vehicle lists which are already immnutable and allocated to a vehicle.
Every shipment taken out the the pool (as time goes forward), when it moves into a vehicle list, may drag others with it, to have a clean cut where the vehicle is delivering to an empty load.
Then I move this's vehicles start location and time_window where it can start as empty vehicle in next pool problems/solutions.

That clipping horizon between pooled shipments, and vehicle-pinned shipments, is very critical, and prone to errors.
This can be helped by multiple vroom calls, where I to both: setting/unsetting the vehicle step-order, to help vroom finding a solution.

Bottom line is, with multiple vroom-calls per incoming shipment, I can use vroom to let it do what it does great.
Thanks for helping!

[jcoupey]

@mbachem-rvk looks like you got some food for though in previous threads and from your own experiments. Shall we close this ticket?

[mbachem]

Hi @jcoupey : yes, thanks, just close it.
If you want to investigate the "skill" issue, just let me know if I can help.

[jcoupey]

OK, closing now. If you think there is a "skill" issue, then maybe worth opening another dedicated ticket?