Abstract
In this paper, we consider the construction of the adaptive control system for linear systems in the self-tuning regulator (STR) type. First, we hypothesize the various change models and then based on them parameters are estimated using the maximum likelihood method. We calculate a posteriori probabilities for all the hypothesized change models using the Bayesian theorem. The model which has the highest probability is considered to be the actual change model. Then according to our regulator design the estimated parameters of the highest probability model are used in optimal control law obtained by dynamic programming and state estimate performed by Kalman filter.
Secondly, we introduce the concept of test signal to improve the system identification.On the other hand, addition of the test signal also results in the disturbance of the system.Thus, in order to have a suitable compromise between the benefit of the system identification and disturbance, we propose to decide the appropriate amount of the test signal.
Finally, we verify in the digital simulation that the test signal input improves the performance index.
