On Performance Evaluation of Inertial Navigation Systems: the Case of Stochastic Calibration

• DOI: 10.36227/techrxiv.20194310
• Publication Date: 2022-07-05T19:12:32Z
• Authors: Davide A. Cucci; Stéphane Guerrier; Mehran Khagani; Lionel Voirol
• Keywords: Robustness

<p>In this work we address the problem of rigorously evaluating performances a inertial navigation system under design in presence multiple alternative choices. We introduce framework based on Monte-Carlo simulations which standard extended Kalman filter is coupled with realistic and user-configurable noise generation mechanisms attempts to recover reference trajectory from noisy measurements. The evaluation several statistical metrics solution, aggregated over hundreds realizations, gives reasonable estimate expected real-world conditions allow user operate choice between setups. To show generality our approach, consider an example application stochastic calibration. Two competing modeling techniques, namely, widely popular Allan variance linear regression, emerging generalised method wavelet moments are compared terms defined scenarios. find that latter provides substantial advantages should be preferred, at least for certain classes sensors. Our allows wide range problems related quantification such as, example, robustness INS respect outliers or other imperfections. While real world experiments essential assess performance new methods they tend costly typically unable lead sufficient number replicates evaluate, correctness estimated uncertainty. Therefore, can bridge gap these pure consideration as done, calibration literature. </p>