Kinematic state estimation of air-borne targets using frequency-weighted Kalman filter aided by artificial neural networks

Estimation of target kinematic state by using a frequency-weighted Kalman filter (FWKF) to obviate the effect of high frequency noise components in estimates has been reported in Bhattacharya, S et al. (2003), Ananthasayanam, MR et al. (2003). However, this introduces phase lag in estimates, which may jeopardize the stability of the guidance system. Again, a target-tracking algorithm employing an artificial neural network (ANN) in cascade with a standard Kalman filter (SKF) has recently been presented in Bhattacharya, S et al. (2003), Bhattacharya, S et al., (2004). It has been shown that the proposed KF-ANN algorithm is promising in improving the quality of estimates without introducing any appreciable lag in the estimates. In the present paper, a synergic approach of KF-ANN and FWKF has been proposed, whereby the estimates from FWKF are post-processed by employing an appropriately trained ANN. The comparative results of SKF, FWKF and FWKF-ANN show remarkable improvement in the proposed FWKF-ANN. The estimates from FWKF-ANN in one hand have reduced high frequency error as compared to KF-ANN, and on the other hand do not contribute significant lag in the estimates as in FWKF, and thus fulfills the desired properties of an estimation algorithm required by advanced guidance system.