Detection and Filtering of Short-Term (1/f) Noise in Inertial Sensors

Combining INS and GPS data by means of Kalman filtering only suppresses sensor noise in a limited frequency band. This paper describes a filtering methodology that allows optimal elimination of the inertial sensor noise over the whole frequency band of interest. It accommodates a model for short-term inertial errors that is especially applicable to ring-laser and fiber optic gyro technologies. It presents a technique for adaptive estimation of the model parameters based solely on data analysis above the edge of the frequency band corresponding to the motion. Using this technique, the model parameters are estimated for a specific system and time period, the noise distribution is predicted within the band of interest and the inertial measurements are de-noised in the wavelet domain. This methodology complements the standard approach of Kalman filtering/smoothing. Empirical testing using different navigation-grade strapdown inertial systems shows an improvement of 20% in attitude determination under dynamic conditions. INTRODUCTION Strapdown Inertial Navigation Systems (INS) provide relative position and orientation at high data rates; typically 50 Hz or higher. If operated in unaided mode, their accuracy degrades quickly due to time dependent errors in their principal observables: angular velocity and acceleration [1]. As the technologies that enable the measurement of these observables keep changing, so do the essential error sources of the measurements. However, the principles for preventing the unbounded growth of INS navigation errors remain quite similar across the broad range of strapdown technologies; the behavior of the main error sources or their combination is modeled and the model parameters are estimated by combining inertial data with some other source of navigation information. Currently, observations using Global Navigation Satellite Systems (GNSS) such as GPS or GLONASS provide the best aiding data for an INS in terms of precision, range and efficiency.