Nonlinear MPC using fmincon. Very slow

3 visualizaciones (últimos 30 días)
MC
MC el 19 de Abr. de 2017
Hello!
I've set up a nonlinear MPC problem with a nonlinear model. The model has 7 states and 5 inputs. Doing a "Run and time" I see that almost all of the used time is spent in fmincon. With a short horizon, N=3 or so, it runs very fast. However as I increase the horizon the problem runs extremely slow. The controller shall track a geometric path with a lenght of about 7 meters at an avg speed of about 1.5 m/s. This run takes about 60 minutes to simulate with a horizon of N=30. It is absolutely essential to have a long horizon of 30 to 40. I've tried both sqp and active set methods.
So I'm wondering if there is a better solver that you would recommend for a nonlinear MPC problem? Or is this the computation time I should be expecting? There is always the option of linearizing, but it is a hassle as the costfunction and one constraint is nonlinear.

Respuestas (0)

Categorías

Más información sobre Model Predictive Control Toolbox en Help Center y File Exchange.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by