TY - GEN
T1 - Deep Learning-driven Automatic Estimation of Smartphone Installation Angles for Vehicle Navigation
AU - Wang, Jingxian
AU - Ding, Weihao
AU - Cui, Bingbo
AU - Shao, Jianbo
AU - Weng, Duojie
AU - Chen, Wu
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - Currently, smartphones are the first choice for vehicle navigation. Due to the low quality of its embedded Inertial Measurement Unit (IMU), some self-constrained technologies have been developed to reduce the divergence of error in GNSS-denied areas, such as Zero Velocity Update (ZUPT) and Non-Holonomic Constraints (NHC). Their rear wheels are considered as the active position of NHC, while smartphones are usually installed on a holder in the front of the vehicle to guide drivers. To ensure the effectiveness of NHC, there is an urgent need to calibrate the lever arms and the installation angles between the smartphone and the previously mentioned active position. The lever arm is relatively stable under most situations since the position of the phone holder in the vehicle is usually fixed, which can be measured by a tape measure or estimated by the parameters of the car directly. The installation angle is difficult to be accurately measured and it may change every time we install the smartphone into the holder. Excluding the roll angle that does not affect the validity of NHC, an automatic estimation algorithm of the pitch and heading installation angles is needed. In this paper, we proposed a deep learning-driven automatic estimation of smartphone installation angles to enhance the performance of smartphone-based vehicle navigation in GNSS-denied areas. In the first step, an Extended Kalman Filter (EKF) is used to integrate GNSS/IμBarometer/DeepOdometry to provide accurate positions and attitudes. Simultaneously, the data of IMU and barometer are input into the trained deep learning network to output the predicted positions with the attitudes obtained from the integrated system. Then, the installation angles are estimated as states in another EKF by differing the predicted positions and the integrated positions. Extensive experiments show that our proposed method can estimate pitch and heading installation angles in deviation within 1 degree.
AB - Currently, smartphones are the first choice for vehicle navigation. Due to the low quality of its embedded Inertial Measurement Unit (IMU), some self-constrained technologies have been developed to reduce the divergence of error in GNSS-denied areas, such as Zero Velocity Update (ZUPT) and Non-Holonomic Constraints (NHC). Their rear wheels are considered as the active position of NHC, while smartphones are usually installed on a holder in the front of the vehicle to guide drivers. To ensure the effectiveness of NHC, there is an urgent need to calibrate the lever arms and the installation angles between the smartphone and the previously mentioned active position. The lever arm is relatively stable under most situations since the position of the phone holder in the vehicle is usually fixed, which can be measured by a tape measure or estimated by the parameters of the car directly. The installation angle is difficult to be accurately measured and it may change every time we install the smartphone into the holder. Excluding the roll angle that does not affect the validity of NHC, an automatic estimation algorithm of the pitch and heading installation angles is needed. In this paper, we proposed a deep learning-driven automatic estimation of smartphone installation angles to enhance the performance of smartphone-based vehicle navigation in GNSS-denied areas. In the first step, an Extended Kalman Filter (EKF) is used to integrate GNSS/IμBarometer/DeepOdometry to provide accurate positions and attitudes. Simultaneously, the data of IMU and barometer are input into the trained deep learning network to output the predicted positions with the attitudes obtained from the integrated system. Then, the installation angles are estimated as states in another EKF by differing the predicted positions and the integrated positions. Extensive experiments show that our proposed method can estimate pitch and heading installation angles in deviation within 1 degree.
KW - Barometer
KW - Installation angles
KW - Smartphone
KW - Vehicle navigation
UR - http://www.scopus.com/inward/record.url?scp=85162917661&partnerID=8YFLogxK
U2 - 10.1109/PLANS53410.2023.10140107
DO - 10.1109/PLANS53410.2023.10140107
M3 - Conference article published in proceeding or book
AN - SCOPUS:85162917661
T3 - 2023 IEEE/ION Position, Location and Navigation Symposium, PLANS 2023
SP - 137
EP - 142
BT - 2023 IEEE/ION Position, Location and Navigation Symposium, PLANS 2023
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2023 IEEE/ION Position, Location and Navigation Symposium, PLANS 2023
Y2 - 24 April 2023 through 27 April 2023
ER -