Temporal Instability
Variations in the zero-point offset of a sensor that change their statistical properties over time represent a persistent challenge for long-term navigation. In the case of non-stationary bias drift, the mean and variance of the error are not constant. This makes the error difficult to remove using simple averaging or static filters.
Driving Influence
Environmental aging and cumulative mechanical stress contribute to this instability. While stationary noise can be characterized by a power spectral density, non-stationary bias drift follows a random walk or a flicker noise model. It causes the error to grow indefinitely if not corrected.
Mitigation Technique
Inertial systems often use external aiding sources like GPS to estimate and correct for non-stationary bias drift. The estimation filter treats the bias as a time-varying state. High-grade sensors are tested using Allan variance plots to identify the different noise components.
This analysis helps engineers understand the time scales over which the bias remains predictable. When a sensor is used in a GPS-denied environment, this drift determines the rate at which the position error accumulates. Compensation algorithms must be updated frequently to keep the drift within acceptable bounds.
Device Grading
Sensors are graded based on the stability of their bias over specified intervals. Minimizing non-stationary bias drift is the primary goal in the development of navigation-grade gyroscopes.
