A New Technique for INS/GNSS Attitude and Parameter Estimation Using Online Optimization

Abstract

Abstract-Integration of inertial navigation system (INS) and global navigation satellite system (GNSS) is usually implemented in engineering applications by way of Kalman-like filtering. This form of INS/GNSS integration is prone to attitude initialization failure, especially when the host vehicle is moving freely. This paper proposes an online constrained-optimization method to simultaneously estimate the attitude and other related parameters including GNSS antenna's lever arm and inertial sensor biases. This new technique benefits from self-initialization in which no prior attitude or sensor measurement noise information is required. Numerical results are reported to validate its effectiveness and prospect in high accurate INS/GNSS applications.Index Terms-Inertial navigation, satellite navigation, velocity integration formula, lever arm, online Reference frame unification is a pre-condition for any accurate information blending of multiple sensors.This work was supported in part by the Fok Ying Tung Foundation (131061), National Natural Science Foundation of China (61174002), the Foundation for the Author of National Excellent Doctoral Dissertation of People's Republic of China (FANEDD 200897) and Program for New Century Excellent Talents in University (NCET-10-0900). A New Technique for INS/GNSS Attitude and Parameter Estimation Using Online Optimization 2For the case of INS/GNSS integration, there is an unavoidable displacement, usually called as the lever arm, between their respective sensing points of INS and GNSS [1]. The magnitude of the lever arm is significant for many applications. For example, when the GNSS antenna is mounted outside on the roof to better receive satellite signals, the INS is usually installed inside the carrier for reasons like easy maintenance. It is difficult or impossible, however, to precisely measure the lever arm vector in a designated frame, taking account of the virtual attributes of the sensing points and the frame axes [2,3]. Lever arm uncertainty acts as a major error source of INS/GNSS in accurate applications, e.g., pose determination in airborne direct georeferencing [4] and airborne gravity measuring [5].A good practice is to accommodate and estimate the lever arm within the extended Kalman filter (EKF) in INS/GNSS integration [2,6,7]. In addition to the knowledge of noise statistics, the Kalman filtering method relies heavily on a roughly known initial attitude that is only achievable under benign situations such as when the carrier is stationary or moving straight. If not properly initiated, however, INS/GNSS integration based on Kalman filtering would be subjected to failure [1,8,9]. These situations are not uncommon in practice, for example when the carrier is moving freely or the duration of benign situations is not long enough to reach a good initial attitude alignment. That is to say, the lever arm estimation is conditioned on a good initial attitude.On the other hand, to obtain a good initial attitude inversely requires a small INS/GNSS lever arm effect. ...

On Misalignment Between Magnetometer and Inertial Sensors

“…Following the technique proposed in the work [23], we can re-express the entities of the Jacobian vector and Hessian matrix in a recursive way, namely, to take the estimation parameters out of the summation. For the sake of brevity, the development and the resultant recursive algorithm is presented in Appendix B.…”

Optimization-based alignment for strapdown inertial navigation system: comparison and extension

“…52, NO. 4 AUGUST 2016 40 Wu et al [23] propose an online constrained-optimization method to simultaneously estimate the attitude and the inertial sensor biases. This online constrained optimization method performs quite well in estimating the attitude and the accelerometer biases.…”

Angle Measurement of Objects outside the Linear Field of View of a Strapdown Semi-Active Laser Seeker

“…Our method needs the following data: the object position $O_T$ (longitude ${\lambda }_T$, latitude $L_T$ and height $h_T$), which is given in advance; the position of the seeker at $t_0$ (longitude ${\lambda }_0$, latitude $L_0$ and height $h_0$); the attitude data of the seeker with the yaw-pitch-roll rotation order at time $t_0$ (yaw angle ${\phi }_0$, pitch angle ${\psi }_0$ and roll angle ${\gamma }_0$) or the quaternions at $t_0$($q_{0\_0},q_{1\_0},q_{2\_0},q_{3\_0}$); the position of the seeker at $t_1$ (longitude ${\lambda }_1$, latitude $L_1$ and height $h_1$); and theattitude data of the seeker with the yaw-pitch-roll rotation order at time $t_1$ (yaw angle ${\phi }_1$, pitch angle ${\psi }_1$ and roll angle ${\gamma }_1$) or the quaternions at $t_1$($q_{0\_1},q_{1\_1},q_{2\_1},q_{3\_1}$). The above positions and attitude data can be obtained via GPS and INS [23,24]. The problem of time consistency can be solved by precisely aligning the above data with the corresponding time.…”