Multi-Sensor Integrated Filtering for Highly Dynamic System Using Recursive Moving Horizon Estimation Technique

This paper presents recursive moving horizon estimation for integrated filtering i.e. estimation of the airborne system navigation parameters. The data fusion/integrated filtering was performed for commercial grade inertial navigation system (INS), global position system (GPS) and air data system. The recursive sliding window discrete time state estimation technique, designed for a non-linear highly dynamic aerial vehicle is executed by minimizing the deterministic cost function. The system's non-linear error and observation models are computed and quaternion based rotation matrix is derived for expressing the attitude angles and inter-coordinate frame transformations. The implemented high order integration scheme avoids the accumulation of errors and high order Butterworth filtering removes the undesirable high-frequency noise. The overall algorithm is tested offline with the data collected from the highly dynamic aerial vehicle flight experiment. The mat-lab environment is used to replicate the noise conditions and implementing the estimation algorithms i.e. MHE and reference Kalman filter. This paper experimental result intimates that proposed moving horizon state estimator is better than conventional estimator as analyzed in performance perspective. Furthermore; it exhibits more stability and resilience in failure mode analysis, as viewed with respect to the onboard GPS and offline conventional filter response.