Browse Prior Art Database

Method for Estimation of Multiple Sensor Bias Components with Limited State Space Dimension

IP.com Disclosure Number: IPCOM000246272D
Publication Date: 2016-May-23
Document File: 7 page(s) / 300K

Publishing Venue

The IP.com Prior Art Database

Related People

Zdenek Kana: INVENTOR [+4]

Abstract

Accurate modeling of the sensor measurement errors is one of the most critical tasks in the design of navigation systems. The number of states being estimated by a statistical estimator depends on the specific sensor model. Low cost sensors such as MEMS need separate modeling and estimation of states for static and dynamic biases. The simultaneous (or parallel) estimation can be computationally complex and demanding especially with three dimensional setting of gyroscope and accelerometer where two triplets of states per sensor are required. A method to use a single triplet of the states per sensor to sequentially estimate both the static and dynamic biases has been proposed. Static bias is estimated with the required accuracy using the Kalman filter first and it is held constant for the next steps. Then the filter is reconfigured to estimate the dynamic bias with a short time constant and dynamic bias is estimated. This method reduces the number of states to be modeled hence reduces the computational demand of navigation and integrity monitoring algorithms.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 30% of the total text.

Page 01 of 7

 Method for Estimation of Multiple Sensor Bias Components with Limited State Space Dimension

Zdenek Kana, Jindrich Dunik, Milos Sotak, Radek Baranek ABSTRACT

Accurate modeling of the sensor measurement errors is one of the most critical tasks in the design of navigation systems. The number of states being estimated by a statistical estimator depends on the specific sensor model. Low cost sensors such as MEMS need separate modeling and estimation of states for static and dynamic biases. The simultaneous (or parallel) estimation can be computationally complex and demanding especially with three dimensional setting of gyroscope and accelerometer where two triplets of states per sensor are required. A method to use a single triplet of the states per sensor to sequentially estimate both the static and dynamic biases has been proposed. Static bias is estimated with the required accuracy using the Kalman filter first and it is held constant for the next steps. Then the filter is reconfigured to estimate the dynamic bias with a short time constant and dynamic bias is estimated. This method reduces the number of states to be modeled hence reduces the computational demand of navigation and integrity monitoring algorithms.


1. Introduction

Accuracy of inertial navigation systems (INS) is significantly affected by the error characteristics of the inertial sensors. Advances in the development of micro- electromechanical systems (MEMS) have made possible the fabrication of cheap and small dimension accelerometers and gyroscopes, which are being used in many navigation applications. These sensors are prone to different errors which, if not treated correctly, degrade the accuracy of the navigation systems. Therefore, a suitable

1


Page 02 of 7

modeling of these errors is necessary in order to take them properly into account and improve the system performance.

Typically the measurement error is composed by a white noise component and a time- correlated noise component often referred to as bias. For MEMS-based inertial sensors or altimeters, the time-correlated component is composed of two biases; static bias with a long time constant and large steady-state variance, and dynamic bias with a short time constant (w.r.t. the constant of the static bias) and small steady-state variance.

Modern navigation systems are realized by a statistical filter (e.g., the Kalman filter or extended Kalman filter) processing a priori and a posteriori information. While a posteriori information comes from the measurements, a priori information is represented by the detailed mathematical model of the system and sensors. Current navigation system architectures are therefore based on the separate modelling (and subsequent estimation) of the static and dynamic biases. As a consequence, the dimension of the state estimated by the filter grows significantly with the increasing number of sensors. In particular, if a Kalman filter based navigation system with one three-axes gyr...