Date of this Version
The Kalman filter provides an effective means of estimating the state of a system from noisy measurements given that the system parameters are completely specified. The innovations sequence for a properly specified Kalman filter will be a zero-mean white noise process. However, when the system parameters change with time the Kalman filter will need to be adapted to compensate for the changes. Traditionally this has been accomplished by using nonlinear filtering, parallel Kalman filtering and covariance matching techniques. These methods have produced good results at the expense of large amounts of computational time. Necessary changes in the system parameters become obvious when the innovations sequence is examined.
Fuzzy logic is an attempt to program human experience into control systems by using a simple set of linguistic rules. In recent years, the use of fuzzy logic has been applied to several types of control systems.
In this thesis, an adaptive algorithm which employs fuzzy logic rules is used to adapt the Kalman filter to accommodate changes in the system parameters. The adaptive algorithm examines the innovations sequence and makes the appropriate changes in the Kalman filter model. To illustrate the effectiveness of this approach, a target tracking system which employs an adaptive Kalman filter to estimate target position is designed and tested.