Precise Gyroscope Bias Estimation Based on Kalman Filter and Allan Variance Analysis

Autonomous Vehicles have gained significant popularity in recent years, necessitating highly accurate navigation and control systems. These systems often rely on inertial navigation systems (INS). The performance of an INS is contingent upon the precision of its inertial measurement unit (IMU) in determining position, velocity, and attitude. However, low-cost IMUs are susceptible to substantial noise and uncertainties, resulting in significant errors in stand-alone operations. While deterministic errors in gyroscopes and accelerometers can be calibrated and compensated. The challenge lies in addressing stochastic errors, which require precise and sophisticated modeling. Various methods such as autoregressive modeling, Gauss-Markov process, power spectral density, and Allan variance can be employed to analyze stochastic sensor errors and determine the noise parameters, enhancing navigation system accuracy. This paper focuses on analyzing stochastic errors in the Crossbow NAV440 IMU sensor utilizing the Allan variance technique, widely regarded as the preeminent method for analyzing stochastic sensor errors due to its ability to evaluate signal stability over a range of time scales. Furthermore, the extracted noise parameters are utilized to estimate gyroscope bias. The proposed technique leading to accurate results by applying a Kalman filter (KF).