This work demonstrates the fusion of two concepts in switching systems, namely, hypothesis testing and multiple model adaptive control. A hypothesis test switching method is defined to detect parameter jumps in a stochastic environment and perform model selection. The control of a discrete-time stochastic system with rapidly time-varying parameters is simulated. Hypothesis test switching is compared to performance index switching, the most researched and popular switching method. It is found that the hypothesis test method is unique because it operates optimally, without user adjustment or a priori knowledge of the time-varying conditions and model placement. Furthermore, it provides more accurate switching and lower control error.
In addition, a major modification to the way multiple models are arranged is proposed. The change decouples stability and performance. As a result, stability is proven more easily, previously required assumptions can be relaxed, new switching methods can be applied, and performance increases are simulated using current switching methods.