Fast In-Motion Alignment for LiDAR–Inertial in Challenging Scenarios

The initial alignment of a vehicular Inertial Navigation System (INS) is a critical process, where accuracy and convergence speed represent the primary performance challenges. This paper addresses initial attitude estimation errors resulting from unreliable aiding information and limited observation time in complex environments, such as during Global Navigation Satellite System (GNSS) signal outages or wheel slippage. To mitigate these issues, a novel dynamic initial alignment strategy aided by Light Detection and Ranging (LiDAR) is proposed to overcome the limitations inherent to conventional GNSS- or odometer (OD)-aided methods. Furthermore, to achieve fast, high-precision alignment with sparse observation data, a dynamic alignment model employing chronological optimization is developed. The proposed algorithm is validated through both simulations and real-world vehicle experiments. Results demonstrate that, compared with existing state-of-the-art methods, the proposed strategy exhibits superior environmental adaptability and faster convergence speed under dynamic conditions.