I have an application(10.1 sp1) using the SVRP tool as a GP service from the toolbox to generate a routing solution to pickup and dropoff seniors thru out the city with five buses/vans. One of the buses which as a capacity of 16 seats sometimes go out and pickup one or two seniors return to the senior center and then go back out to pickup 3 or 5 and and drops them off at the senior center again etc. etc. Is their a way to have the bus full or almost full to capacity before making the dropoffs. I am thinking this may have something to do with cost but I am not sure.
I read the documentation and tried a few scenarios but still have a few question. Attached is an output from a SVRP solution. For a particular bus it picks up 6 seniors(all going to the senior center). Two of the seniors have an 8-12 time window everybody else has a 8-11 window. The "solver" drops off only 2 seniors(with 8-12 time windows) and leave the other 4 seniors on board the bus. It then goes out and pick up another 5 seniors going to the same senior center (with 8-11 time window) and then drops off all 9 seniors at the center. Why didn't the "solver" drop-off the first 6 seniors at the center to begin with (instead of 2). What role does the time windows plays when calculating a solution? Note: all seniors have a max travel time of 40 mins.
Update: I edit the time window for the 2 seniors from 8-12 to 8-11 and still got the same results
Looking at your result, perhaps the route does not have enough time to complete all of the deliveries and meet the remaining time windows. Looks like you may have a one minute service time on the delivery orders?
One thing you can do is make use of the ArriveDepartDelay property on the routes to more accurately account for loading and unloading groups of people. We added this to better model unloading a bus. For example, it might take 5 minutes to unload 1 kid and 6 minutes to unload 10. So an ArriveDepartDelay of 5 minutes on the route and perhaps .25 minutes service time on the order models it better. This property also has the effect of not making it "free" to stop the route to pick up some orders because it is already passing by that location. In other words, better stop grouping.
Thanks for your input. I tried your suggestion of 3 min ArriveDepartDelay, 0.25 Service Time (along with some others) and it seems to improve the grouping slightly but I still get some output that still baffles me. The first jpg show the SVRP solution to go to the same pickup location three time to dropoff three seniors at different locations. The only difference that I can find is the Time window for the pickup. Each has three different time slot(see jpg) but they all begin at the same time (2:30). I then modify the time so that all three starts and end at the same time (2:30-3:30) and ran the solution. This time the SVRP had them all pickup at the same time (see second jpg). It seems that the solver favors the later part of a time window
Does the ArriveDepartDelay field looks at the time window in order to get a solution? Note: the pickup and dropoff time are the same for the ordered pair.
Can the solver favor the earlier part of the time window instead of the later part?