抄録
This paper presents the method of evaluating the performance loss and of improving the performance of the Wiener-Kalman filter where identification errors exist in mathematical models of dynamic and observation systems.
The description is mainly divided into two parts. The first part is concerned with the performance loss of the discrete-time Wiener-Kalman filter designed on the basis of the model with errors in both transition and observation matrices. The difference equation which describes the evolution of the actual error covariance matrix is derived. For the first order system, it is shown that the Wiener-Kalman filter is more insensitive to parameter variation in observation model than in system model. The relation between the estimation and the noise free regulator problem is also examined. The second part is devoted to the development of a method of designing linear filter, when the elements of both the transition and observation matrices are randomly time-varying. For this purpose the result of the first part is utilized. Results of digital simulation studies are demonstrated to show the effectiveness of the linear filter presented here.
