Re: [sumo-user] Multiple environments in reinforcement learning

[Michael Behrisch <oss@xxxxxxxxxxx>]

Hi Jordan,
can you give more details on your system? Some systems limit the nummber 
of open sockets or have a large delay on freeing a closed socket 
(especially Windows). If this does not apply here I am afraid you can 
only try to catch the exception and restart the sumo instance possibly 
after some wait.

Best reagrds,
Michael


Am 17.01.23 um 09:25 schrieb Jordan Poots:
> Hi,
>
> I've been using reinforcement learning with sumo and recently passed 
> over to a new system. I keep finding on this new system I have errors 
> where my training collapses due to some sort of collision when trying to 
> open a new environment. Note that this error is not immediate, this 
> particualr one occured over 1.5m training steps in.
>
> Error: tcpip::Socket::accept() Unable to create listening socket: 
> Address already in use
> Quitting (on error).
>
> Is there a way to ensure this doesn't happen?
>
> Thanks,
>
> Jordan
>
> _______________________________________________
> sumo-user mailing list
> sumo-user@xxxxxxxxxxx
> To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/sumo-user