<h2>AAS  96-108</h2><h2>A NEW CLASS OF ADAPTIVE, EXTENDED KALMAN FILTER</h2><h4>D. R. Cruickshank and G. H. Born, University of Colorado</h4><h2> Abstract </h2>A new type of adaptive Extended Kalman Filter (EKF) incorporating genetic adaptation of the process noise matrix is developed. The process noise matrix structure is inherited from the Dynamic Model Compensation (DMC) algorithm. The adaptation process, termed Genetic Model Compensation (GMC), optimizes the process noise terms for acceleration, velocity, position, clock drift, and clock bias. Simulation results show that GMC creates a self-tuning process noise matrix, resulting in an adaptive EKF which achieves accuracies equivalent to an optimized DMC filter but without the need for a priori estimates or tuning of the process noise parameters.<br><br>