An efficient tuning framework for kalman filter parameter optimization using design of experiments and genetic algorithms

The Extended Kalman Filter (EKF) is currently a dominant method of sensor fusion used for navigation of mobile devices, robotics and autonomous vehicles. One navigation system involves the EKF fusion of an Inertial Navigation System (INS) with a Global Navigation Satellite System (GNSS) to perform 3D pose estimation, which is essential to practical applications like autonomous vehicles and UAVs. This system involves using the INS to predict pose changes and uses GNSS measurements when available to correct for estimation drifts and sensor errors. The performance of the state estimation is heavily dependent on the accurate selection of EKF parameters, leading to the optimal selection of parameters being a critical factor in the design and use of the EKF. In this paper, a methodical and efficient method of EKF parameter tuning is presented. The tuning framework uses nominal parameters generated by Gauss Markov (GM) and Allan Variance (AV) methods that are tuned by Genetic Algorithms (GA) accelerated by a Design of Experiment (DoE) technique to efficiently optimize EKF parameters. This framework has been implemented in MATLAB and tested using simulations and real data. The results demonstrate that GA tuned parameters increases accuracy substantially, and that the DoE technique consistently improves the convergence behavior of the GA