Hi,

I am using vroom and my solutions are often sub optimal and I can spot right away a better solution if I look at a map.
I am using vroom with 1000-1500 jobs and 30 vehicles. But my time windows are such that usually only 5 jobs can be assigned to a vehicles. I also have priorities among this jobs, priorities going from 1 to 5.
The solutions usually becomes better once I limit the number of vehicles of solutions. But this is not doable for my real world solution.

I am reaching some limits of the optimisation or I am doing something wrong?
By suboptimal I mean

• Sometimes there is a doable solution that can have a higher sum of priorities
• Sometimes for the same number of points there is a shorter solution

I am using vroom 1.5.0.
with this config file:
cliArgs:
geometry: false # retrieve geometry (-g)
threads: 3 # number of threads to use (-t)
explore: 5 # exploration level to use (0..5) (-x)
limit: '5mb' # max request size
logdir: '/..' # the path for the logs relative to ./src
maxlocations: 5000 # max number of jobs/shipments locations
maxvehicles: 500 # max number of vehicles
override: true # allow cli options override (-g, -t and -x)
path: '' # VROOM path (if not in $PATH)
port: 3000 # expressjs port
router: 'osrm' # routing backend (osrm, libosrm or ors)
timeout: 300000 # milli-seconds

I have tested it by putting an input with an other matrix, not computed from my local osrm but I still have problems.

I can provide a reproducible example, but I can't post it here because I don't want to expose my data... (just send me a mail at rom.rey06 at gmail.com and I can send it to you)

Thanks for your help!