27 May 2024 in IEEE Transactions on Vehicular Technology
High accuracy in-motion alignment and integrated navigation with assistant of body-velocity is a nonlinear state estimation problem as existence of attitude. The Extended Kalman filter (EKF) is a frequently used tool but criticized for its inconsistency problem especially with large initial misalignment angle, which hinders the improvement of navigation accuracy. This paper proposes a backtracking in-motion alignment framework using enhanced Lie group (LG-EKF-w) to replace troublesome EKF. The LG-EKF-w captures the nature of 3-dimension rigid body motion, which guarantees the inconsistency and improves the accuracy of state estimation. Land and sea field experiments evaluates the performance of LG-EKF-w compared to EKF, state transformation extended Kalman filter (ST-EKF) and LG-EKF-i. The results show that LG-EKF-w has faster initial alignment speed compared to EKF and better positioning accuracy under badly aligned conditions.
6 January 2026 in Results in Engineering
14 June 2026 in Measurement
4 August 2026 in Measurement Science and Technology
In GNSS-denied environments, the integration of a strapdown inertial navigation system and an odometer (OD) is common for land vehicle autonomous navigation. However, traditional error-state filters may suffer from degraded accuracy during in-motion alignment when starting from large misalignment angles, primarily because the linearization point can deviate from the true state. To address this, this paper proposes a closed-loop supplementary correction algorithm. First, the initial attitude error is estimated in real time within the state transformation extended Kalman filter framework. Through geometric decomposition, the proposed method estimates the navigation error components caused by the initial deviation. Unlike existing open-loop methods, the proposed algorithm implements a closed-loop feedback mechanism that iteratively improves the filter’s linearization point, alleviating residual nonlinear effects and accelerating error convergence. Second, the vehicle position is derived as a linear function of the initial attitude and OD parameters. Using the results of the closed-loop correction, the position can be obtained directly without storing and reprocessing the historical sensor data. Simulation and field experiment results demonstrate that the proposed method can improve attitude and position estimation accuracy during in-motion alignment. It provides an efficient and practical initial alignment solution for vehicle-borne integrated navigation systems, showing potential for engineering applications.