GNSF exact Hessian


First some advertisement for the GNSF type integrators. In my specific problem (Gauss-Newton SQP for tracking MPC (horizon N = 20) for a - quite nonlinear - airborne wind energy system) I get a speed-up of factor 2.37 in time per SQP iteration and a factor 3.73 in time spent in linearization. w.r.t. IRK. Impressive!

However, it seems that the exact Hessian option is currently not supported for GNSF integrators. This could make a huge impact on e.g. economic MPC performance. Is this feature somewhere on the horizon?


1 Like

Hi Jochem,

great to hear GNSF is performing so well on your problem.

I still have this on my to do list and looked a bit into it, but it is not very straight forward.
I hope I can tackle it again early next year.