Model-Free Learning for Massive MIMO Systems: Stochastic Approximation Adjoint Iterative Learning Control

Learning can substantially increase the performance of control systems that perform repeating tasks. The aim this letter is to develop an efficient iterative learning algorithm for MIMO with a large number inputs and outputs does not require model knowledge. gradient criterion obtained through dedicated experiments on system. Using judiciously selected randomization technique, unbiased estimate from single experiment, resulting in fast convergence Robbins-Monro type stochastic descent algorithm. Analysis shows approach superior earlier deterministic approaches related SPSA-type algorithms. illustrated multivariable example.