Article
An Improved Yaw Estimation Algorithm for Land
Vehicles Using MARG Sensors
Gang Shi
1,2,3
, Xisheng Li
1,4,
* and Zhengfu Jiang
1
1
School of Automation and Electrical Engineering, University of Science and Technology Beijing,
Beijing 100083, China; shigang_upc@163.com (G.S.); jzf_ustb@163.com (Z.J.)
2
College of Information and Control Engineering, China University of Petroleum, Qingdao 266580, China
3
Shengli College, China University of Petroleum, Dongying 257061, China
4
Beijing Engineering Research Center of Industrial Spectrum Imaging, Beijing 10083, China
* Correspondence: lxs@ustb.edu.cn; Tel.: +86-010-6233-4885
Received: 8 August 2018; Accepted: 25 September 2018; Published: 27 September 2018
Abstract:
This paper presents a linear Kalman filter for yaw estimation of land vehicles using
magnetic angular rate and gravity (MARG) sensors. A gyroscope measurement update depending
on the vehicle status and constraining yaw estimation is introduced. To determine the vehicle status,
the correlations between outputs from different sensors are analyzed based on the vehicle kinematic
model and Coriolis theorem, and a vehicle status marker is constructed. In addition, a two-step
measurement update method is designed. The method treats the magnetometer measurement update
separately after the other updates and eliminates its impact on attitude estimation. The performances
of the proposed algorithm are tested in experiments and the results show that: the introduced
measurement update is an effective supplement to the magnetometer measurement update in
magnetically disturbed environments; the two-step measurement update method makes attitude
estimation immune to errors induced by magnetometer measurement update, and the proposed
algorithm provides more reliable yaw estimation for land vehicles than the conventional algorithm.
Keywords:
attitude estimation; Kalman filter; land vehicle; magnetic angular rate and gravity
(MARG) sensor; quaternion; yaw estimation
1. Introduction
As a set of Euler angles, the yaw, pitch, and roll represent the orientation of a body frame with respect
to a reference frame. The pitch and roll are also referred to as attitude. Yaw and attitude estimation are
widely used in vehicular technologies including driver assistance [
1
–
3
], vehicle safety [
4
,
5
], etc. In recent
years, magnetic angular rate and gravity (MARG) sensors [
6
] are widely used in orientation estimation.
A MARG sensor consists of a triaxis magnetometer, a triaxis gyroscope, and a triaxis accelerometer.
Reasonable installation and calibration make it acceptable to assume that the sensor frames are aligned
with the body frame. Hence, a MARG sensor can measure the geomagnetic field, angular rate of the body
frame, and the gravity resolved in the body frame in undisturbed environments.
In order to obtain an orientation estimation of the body frame, we can integrate the gyroscope
output based on an initial value, but the result will drift away with time because of gyroscope
measurement errors [
7
]. Alternatively, we can solve the Wahba problem [
8
] using magnetometer and
accelerometer outputs, but the sensor outputs are apt to be interfered by motion accelerations and
magnetic disturbances [
9
,
10
]. Therefore, to make the most of the information from MARG sensors and
obtain robust orientation estimation, many fusion algorithms have been studied. These algorithms
can be classified into two categories: one is based on complementary filters, which realize the
fusion in frequency domain [
11
–
13
], and the other is based on Kalman filters, which employ a
Sensors 2018, 18, 3251; doi:10.3390/s18103251 www.mdpi.com/journal/sensors
