unexpected number of optimization variables in Ipopt about towr HOT 2 CLOSED

Yes, good observation. This difference comes since IPOPT doesn't regard variables with same upper and lower bound as variables anymore, but just fixes their value.

In towr the motion-splines are parameterized by nodes, but then various nodes in this spline are fixed in advanced, such as initial pos/vel/footholts, final, etc. Other elements that are fixed are those coming from the phase-based endeffector parameterization, and also don't show up in IPOPT anymore, e.g.

Of course we could also remove the variables from the optimization problem instead of setting equal upper and lower bound. However, using bounds and always adding the full spline as variables we found makes code/formulation more extensible and elegant.

from towr.

Great. I did not expect Ipopt's preprocessing to be so "clever". I agree that it makes perfectly sense to keep the fixed variables for a more elegant formulation of the problem. Especially considered that Ipopt takes care of that and therefore the computation time should not be affected.

from towr.