fmincon has 4 algorithm options: 'interior-point', 'trust-region-reflective', 'sqp', 'active-set'. Is OptimTraj capable of using any of these 4 NLP algorithms? Does OptimTraj default to use the default fmincon algorithm 'interior-point'? about optimtraj HOT 1 CLOSED

Good Question!

The default options for all OptimTraj problems are set in getDefaultOptions.m. Right now, they always select the default FMINCON solver, which is interior-point. This is intentional.

Let's look at the FMINCON page on algorithm selection. There are four algorithms. Of these algorithms, sqp and active-set are not good for trajectory optimization, as they are only for small and medium scale problems. Trajectory optimization creates large sparse problems. Of the remaining two algorithms, the trust-region-reflective requires a gradient, and only works for some constraints. Thus, it cannot handle all problems coming from OptimTraj. The final solver, interior-point, handles all problems, and thus is the most desirable choice.

from optimtraj.