Bayesian Statistical Analysis for Performance Evaluation in Real-Time Control Systems

This paper presents a method for statistical analysis of hybrid systems affected by stochastic disturbances, such as random computation and communication delays. The method is applied to the analysis of a computer controlled digital hydraulic power management system, where such effects are present. Bayesian inference is used to perform parameter estimation and we use hypothesis testing based on Bayes factors to compare properties of different variants of the system to assess the impact of different random disturbances. The key idea is to use sequential sampling to generate only as many samples from the models as needed to achieve desired confidence in the result.