In use parameter estimation of inertial sensors by detecting multilevel quasi-static states

Abstract

We present an autoadaptive algorithm for in-use parameter estimation of MEMS inertial accelerometers and gyros1 using multi-level quasi-static states for greater accuracy and reliability. Multi-level quasi-static states are detected robustly using data from both gyros and accelerometers. Proper estimation of time-varying sensor parameters allows us to develop a mixed-reality real-time hand-held orientation tracker with dynamic accuracy of less than 20. Existing methods like Kalman filters do not take time-varying nature of parameters into account, instead modelling the time-variation as higher values in noise covariance matrices; thus underestimating the sensor capabilities. © Springer-Verlag Berlin Heidelberg 2005.