[Ipopt] Infeasible problem detected during optimization

[Stefan Vigerske]

Hi,

I would guess that Ipopt claims infeasibility if it gets to point where 
it cannot find any direction that decreases feasibility.
Restarting Ipopt from a solution close to the variable bounds results in 
moving that initial point, so Ipopt would not start from the locally 
infeasible point in your restart. Also the Barrier parameter mu is reset 
when restarting, so Ipopt when restarted is likely to behave differently 
than when it got stuck.

Stefan

On 04/30/2018 10:15 PM, Maxime Boulay wrote:
> Hello,
> 
>   
> 
> I have a problem where IpOpt initially arrives to a status of infeasible
> problem, but after starting it again with a warm start he arrives very
> quickly to the right solution. So I am wondering, what are the criterias for
> IpOpt to consider the problem as infeasible? If it comes to a local point of
> infeasibility, is it possible to make it continue by itself automatically so
> that it gets to the right solution without needing a warm start?
> 
>   
> 
> Thank you,
> 
>   
> 
> Maxime
> 
>   
> 
> 
> 
> 
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> https://list.coin-or.org/mailman/listinfo/ipopt
>