Performance of mixtures of adaptive controllers based on Markov chains

When dealing with nonlinear non-Gaussian systems, Markov chain seems to be a promising choice. However, the curse of dimensionality inherent to Markov chain restricts applicability of this model. This motivates the search for a feasible approximation. This paper proposes an approximate adaptive control methodology for large-state-space and high-order Markov chains. Moreover, various experiments are presented, which further confirm the feasibility of this methodology