Soft decision fusion using kalman filter and fuzzy logic

The applications of Kalman filter include tracking, navigation, guidance, control and parameter estimation. The performance of the filter depends on how accurately the mathematical models for actual dynamic system and measurement device are known, and also on selection of tuning parameters such as process noise covariance (Q) and measurement noise covariance matrices (R). Relying on tuning parameters selected through trial and error approach, especially Q often compensates the modelling errors. This is not an easy task for a filter designer and becomes an extra burden in terms of processing time involved. In such situations, Fuzzy logic concept can be applied to tackle the tuning aspects. Fuzzy logic uses an intuitive experience based-approach for problems that are too difficult to tune properly. This report tries to combine Fuzzy logic and13; Kalman filter for filtering (Fuzzy Kalman filter/FKF) and sensor data fusion problems for target tracking application. The performance is evaluated using numerical simulation examples. The approach of this report is compared with two existing approaches of state vector fusion i) one based on covariance weighting factors for fusion and ii) the other with weights determined empirically from the innovations.