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

Block diagram of tuning framework

EKF Block Diagram

Standard GA results

Hardware setup

Tuning Trajectory

Simulated (left) and Real (right) nominal results with error plots

Multiple GA runs for simulations and real data

Simulated (left) and Real (right) GA results with error plots, comparison with Nominal results

DoE parameter impact results

Multiple DoE enhanced GA runs for simulated and real data

Simulated (left) and Real Data (right) GA with DoE results with error plots, comparison with GA results

IMU biases convergence to the true bias values in the simulated IMU data

Trajectories used

Tuning dataset (red), Test set #1 (yellow), and Test set #2 (blue) overlaid on Google Maps

Pose Errors for Test set #1

Pose Errors for Test set #2

Innovations for Tuning Dataset

Innovations for Test set #1

Innovations for Test set #2

IMU Bias for Tuning Dataset

IMU Biases for Test set #1

IMU Biases for Test set #2

Outage Test for Test set #1

DoE results on varying parameter configurations

RMSE vs 6 selected parameters for 50 randomized parameter configurations