Relevance of transport depot sequence before optimization

[Oliver Meyer]

Dear PTV team,

we have a question regarding the XTour optimization API that we use in our product:

Does the optimization re-arrange the order of transport depots in an overnight tour at will (only considering the most efficient tour), or does it somehow try to retain the sequence as long as it doesn't interfere with the efficiency of the tour?
Will the optimization also freely distribute transport depots between days on a multi-day tour?

I ask this because some of our customers feel that the routes we plan for them by xTour don't seem to be optimal, and we suspect that this might have something to do with the sequence of the transport depots in the input tour.

Thanks in advance.

[Bernd Welter]

Hello Oliver,
let me give you some basic statements about the optimization. In the end it's all about

• select a subset of the given orders (best case: all)
• try to arrange them in a sequence that performs optimal (depending on the goal)
• and ensure that the constraints are not violated (such as capacities, times and skills)

For us there is no need to keep an existing sequence - we evaluate moves right from the beginning.
And depending on the visiting patterns we evaluate a number of independent scenarios via the overnight stays (0 stays, 1 stays, 2 stays, ...) and then pick the overall optimal result. This is the reason why an overnight planning might sometimes take longer than a (more or less) simple sequencing.

I would be interested in a concrete example of your customer. Usually we ask in return: what is the sequence you'd expect to receive? Did you already try to set this sequence as an input tour to check whether it is free of violations?

Best regards
Bernd

[Bernd Welter]

And here is some important feedback from DEV: compact as hell!

• optimization means change towards a better solution (tour period, driving period, distance)
• input sequence has to valid (even considering the time windows of vehicle and stops)
• an input tour is optional in this usecase but does not guarantee a better result

Thanks Frank for the additional comments,

Bernd Welter

[Oliver Meyer]

Hello,

thanks for your reply.

I just would like to elaborate on the background of my question:

When we submit a request, the depots are more or less in a random order.

Now one of our customers has observed that sometimes the tour considered to be optimized would look rather weird to them; for example, given some first station, the algorithm would sometimes select a far-off station over a nearby one for second station, thus increasing the driving distance while reducing the number of stations that could be considered in the given time frame, which is not what our customers want.

We first thought that this could have something to do with the sequence of stations in the input. If xTour would have tried to keep the sequence as much as possible, this would have explained these results.

I've filed a ticket for clarification with your help desk (#PTV199169), and the response was that this is due to the MergeSavings algorithm. The response also more or less implied that there was nothing we could do about this when using xServer as SaaS.

You can find an example request attached to the ticket if you want to check for yourself.

Regards
Oliver Meyer

[Bernd Welter]

Hello Mr. Meyer,

how about the approach described in this post?
http://xserver.ptvgroup.com/forum/viewt ... ?f=6&t=403

I applied the prio 5 to the 5 scheduled oreders and prio 6 to the others.
With that input the server schedules 7 instead of 5 orders.

Best regards,
Bernd