INS and INS/GPS Accuracy Improvement Using Autoregressive (AR) Modeling of INS Sensor Errors

摘要: Current Inertial Navigation System (INS) error models that are used in most INS and INS/DGPS applications have some limitations, which turn affect the overall system accuracy. One of these limitations is associated with stochastic modeling sensor errors. For navigation-grade IMUs (gyro drift 0.005-0.01 deg/h), a 1st order Gauss-Markov (GM) model usually for this purpose. This also true low-cost inertial systems 100-1000 deg/h) although sometimes white noise process utilized instead. Using long records static data, collected by IMU, we showed actual Autocorrelation Sequence (ACS) such sensors not always well represented GM process, where computed ACSs higher-order terms. The approximation processes can lead to significant accuracy degradation. To overcome problem, paper discusses errors using Autoregressive (AR) orders higher than one. Modeling AR was initially tested on limited data set. Initial results performance much better commonly process. Based results, been extensively two kinematic tests land vehicle environment. In paper, behavior different categories (navigation-grade, medium-grade low- cost) investigated. Numerical analyses performed illustrate poor obtained from experimental data. Finally, presented analyzed case stand-alone navigation positioning during frequent DGPS outage situations.