A two-input two-output robust multiple model adaptive control (RMMAC) case study

Abstract

We use the RMMAC architecture and design methodology, introduced in the work of M. Athans et al. (2005) and S. Fekri et al. (2004), to design and evaluate a truly multivariable adaptive control system; this fills a void that is present in the literature. A three-cart problem with two uncertain parameters, two controls, and two outputs is used as the case study. We show that the RMMAC significantly improves disturbance-rejection compared with the "best" non-adaptive controller designed by mixed-mu synthesis. In the example considered, in addition to two uncertainties in mass and spring constants, there are unmodeled dynamics as well as (unmeasured) stochastic disturbance inputs and noisy sensor measurements. Numerous simulation results are presented that demonstrate the superior performance of the RMMAC