You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Many (too many !) fields were added in the DOCP that complete / mirror the ones in OCP, partly for convenience and also to fix some types. Once OCP is redone, DOCP should be updated, with 2 possible directions:
a) keep a copy of the OCP in DOCP, and remove all the duplicate fields
b) remove the OCP and make DOCP self-sufficient, with a better constructor taking the ocp and the few discretization parameters, instead of all the individual fields
Having the OCP inside is nice since we have all relevant information available, in particular when rebuilding the OCP solution from the discrete one. If we manage to have good types in the OCP I think it would be the easier path. One potential drawback is that it makes DOCP less generic since it ties it to the OCP, however it is not completely clear how we would use the DOCP outside of OptimalControl / CTBase anyway.
Notes:
DOCP is currently non-mutable, which supposedly helps with code optimization
fields could also be reorganized into sub-structs for a more compact constructor call. For instance we already have a struct inside DOCP for the discretization.
The text was updated successfully, but these errors were encountered:
Many (too many !) fields were added in the DOCP that complete / mirror the ones in OCP, partly for convenience and also to fix some types. Once OCP is redone, DOCP should be updated, with 2 possible directions:
a) keep a copy of the OCP in DOCP, and remove all the duplicate fields
b) remove the OCP and make DOCP self-sufficient, with a better constructor taking the ocp and the few discretization parameters, instead of all the individual fields
Having the OCP inside is nice since we have all relevant information available, in particular when rebuilding the OCP solution from the discrete one. If we manage to have good types in the OCP I think it would be the easier path. One potential drawback is that it makes DOCP less generic since it ties it to the OCP, however it is not completely clear how we would use the DOCP outside of OptimalControl / CTBase anyway.
Notes:
The text was updated successfully, but these errors were encountered: