Moving window recursive filtering for integrated navigation performance enhancement

This paper presents an alternate recursive state estimator of moving horizon estimation as compared to the conventional state estimator of Kalman filtering. The state estimator estimates the position set, velocity set and attitude data of the dynamic aerial vehicle. In the first scenario, the available data for processing are the measurement set from accelerometers assembly, gyroscope triad, and global positioning system (GPS) from the low cost inertial measuring unit (IMU). The GPS position and GPS velocity measurements are the aiding source as well the comparison reference of the computed solution of the dynamic aerial vehicle (DAV). In the second scenario receding window discrete time state estimator, extended for non-linear vehicle navigation application is implemented using the deterministic cost function of least squares, for integrated filtering. The receding data window approach uses the past measurements samples over the horizon as the tuning parameter of the moving horizon state estimator. The moving horizon state estimator is evaluated offline in the numerical experiment including the flight testing data collected. The flight test on a small aerial vehicle with all the sensors, GPS receiver, power systems instrumented onboard is considered for this paper. Matlab platform is used to simulate the noise environment and implementing the algorithms. This paper designed algorithm results reveals that proposed moving horizon estimator is faster in convergence in the presence of large initialization errors, linearization errors and outliers as compared to the reference filter i.e. EKF tested online and offline.