Stochastic descent vs. sample average in uncertain minimax control problems

ARAGONE, LAURA S.; GIANATTI, JUSTINA; LOTITO, PABLO A.; PARENTE, LISANDRO A.

Comodoro Rivadavia

Congreso; VI MACI 2017 - VI Congreso de Matemática Aplicada, Computacional e Industrial; 2017

ASAMACI

We propose a stochastic descent algorithm for solving uncertain minimax control problems. At each iterate, the scheme randomly draws a sample point from the underlying probability space, computes a feasible direction and performs an Armijo step in order to obtain the next iterate. On some simple examples, we compare its performance with a descent algorithm based on sample average approximations, obtaining promissory numerical results.