Modeling and bounding low cost inertial sensor errors

This paper presents a methodology for developing models for the post-calibration residual errors of inexpensive inertial sensors in the class normally referred to as "automotive" or "consumer" grade. These sensors are increasingly being used in real-time vehicle navigation and guidance systems. However, manufacturer supplied specification sheets for these sensors seldom provide enough detail to allow constructing the type of error models required for analyzing the performance or assessing the risk associated with navigation and guidance systems. A methodology for generating error models that are accurate and usable in navigation and guidance systems' sensor fusion and risk-analysis algorithms is developed and validated. Use of the error models is demonstrated by a simulation in whichthe performance of an automotive navigation and guidance system is analyzed.