Multi-level localization trajectory alignment and repairing in complex environment

Abstract

At present, multi-device collaborative navigation is popular in some projects. Most existing collaborative navigation algorithms depend on point clouds or 3D maps. However, ordinary users may encounter challenges in accessing raw data due to the security and privacy of spatial data. As a solution, this article introduces the algorithm of multi-device navigation result alignment and repairing at the trajectory level. This work faces three main challenges: (1) it is hard to ensure the availability and overall accuracy of global alignment without point clouds; (2) in a complex campus environment, light detection and ranging (LiDAR) inertial odometer (LIO) results are bound to gradually diverge due to the absence of precise absolute position data for correction, leading to error accumulation in the local trajectory; (3) in LiDAR challenging environments, such as rapid turns, LIO trajectories may exhibit outliers. In response to these three challenges, this article proposes a multi-level (global-local-detail) multi-device localization trajectory alignment and repairing idea and designs the corresponding methods. Specifically, (1) at the global-trajectory level, the reliable GNSS points are used for multi-device trajectory alignment; (2) at the local-trajectory level, a segmented forward–backward smoothing algorithm is applied to enhance local trajectory accuracy; (3) at the detailed-trajectory level, outlier detection and repairing algorithm based on motion mode is employed to ensure the reliability of detailed trajectories. The experiments were conducted in different environments, and the results validated the high accuracy and reliability of the proposed algorithm.

Keywords

• Localization trajectory alignment
• Motion mode recognition
• Segmented forward–backward smoothing
• Outlier detection and repairing
• GNSS