search

MadNLP / DynamicNLPModels.jl

NLPModels for dynamic optimization
MIT License
11 stars 1 forks source link

Hessian bug fix #4

Closed sshin23 closed 2 years ago

sshin23 commented 2 years ago

Normally, MadNLP shouldn't regularize the problem for convex QPs, but it is regularizing this problem. This is an indication of a potential bug. This should be root-caused and fixed https://github.com/MadNLP/DynamicNLPModels.jl/runs/6838395675?check_suite_focus=true#step:5:656

dlcole3 commented 2 years ago

@sshin23, I think the reason for this error is that I have a a $u_N$ variable defined, which results in an empty column in the Jacobian and an empty row and column in the Hessian. There is no need for a $u_N$ decision variable, is there? I don't see it coming up in any of the constraints of the Jerez et al. paper. If so, I will go ahead and remove it.

sshin23 commented 2 years ago

yes, empty variable causes an inssue