Multi-Sensor Attitude and Heading Reference System Design Using Genetically Optimized Kalman Filter

Attitude and Heading Reference System (AHRS) is a self-contained sensors assembly that can estimate full 3D orientation of an object. The AHRS system model involves integration of angular rate measurements from gyroscope which are fused with absolute measurements from magnetometer/accelerometer using Extended Kalman Filter (EKF). EKF accuracy is greatly affected by process noise parameters and measurement noise parameters. Therefore, this thesis developed a systematic method of EKF noise parameters optimization using a hybrid stochastic, Genetic Algorithms (GA)-based approach supported by Design of Experiments (DoE) technique. The proposed approach has been developed in MATLAB and tested on simulation data and verified on real data collected under different scenarios. Results showed that the proposed approach can provide 40-60% better accuracy compared to conventional methods within few GA iterations. In addition, application of DoE technique reduces GA iterations to convergence by approximately 60%.