osqp does not always return the same result eventhough parameters are unchanged about osqp-r HOT 4 CLOSED

The reason is that the adaptive_rho feature is configured by default to update rho and refactor the linear system in the solver based on the relative time taken between the factor phase and the iteration phase of the solver. The factorisation code is deterministic, but the execution time can vary b/c it is competing for CPU resources with other things on the machine.

If you want completely repeatable behaviour, I would suggest configuring the solver the update rho instead after a fixed number of iterations (maybe try 50-100 to start).

from osqp-r.

Hi Paul,

Many thanks Paul for your quick and comprehensive answer!
It works fine on my side. But just to be sure, can you confirm that I only need to add in osqpSettings:
adaptive_rho_interval = 100

Kind regards,

Philippe Carpentier

from osqp-r.

I think that should work, yes. Reopen the issue if not though.

from osqp-r.

Once more, many thanks Paul,

Philippe Carpentier

from osqp-r.