[Coin-ipopt] Re: Coin-ipopt Digest, Vol 14, Issue 2

[Zhiwen Chong]

Hello all,

I'm solving a discretized DAE system with IPOPT 2.2.1e/AMPL, and I keep getting
this error message:

    Constraints seem rank-deficient!

Despite this, the problem solves correctly (OPTIMAL SOLUTION FOUND)... although
sometimes it doesn't.

Note: my system is pre-initialized with results from a simulator, so NLP is
starting from a feasible point. The constraints are also scaled appropriately.

My questions are:
1) Can I safely ignore the above error if IPOPT exits with OPTIMAL SOLUTION
FOUND?
2) Are there any methods/tools for tracking down linear dependencies, especially
in an intermediate iteration? (The standard way of going about this is to do a
QR decomposition, but I cannot for the life of me figure out how to do this in
AMPL.)
3) Is the IPOPT algorithm a feasible path algorithm?

Thanks.