Using a metric based tuning of Converted Measurement Kalman Filter (CMKF) for realistic target tracking scenario

Abstract

Converted measurement Kalman filter (CMKF), which is used in target tracking problems, converts polar measurements into Cartesian form to match Cartesian states. However, filter performance may degrade due to incorrect filter tuning of the biases and nonlinearities that arise due to the conversion. In this paper, the necessary converted measurement noise covariances have been derived in a 6 degree of freedom (6DOF) realistic target tracking scenario. In order to do so, the necessary frame conversions have also been accounted for to enable the estimation of the relative kinematics of the target. Thereafter, existing notions of robustness and sensitivity metrics have been suitably adapted to interpret the performance of CMKF in the 6DOF platform for various choices of tuning parameters and to predict the optimal tuning choice. The predicted performances and the optimal filter tuning choice have been validated in simulations in terms of the root mean squared errors (RMSE) for all measurements for 5000 Monte Carlo runs.