Fast Inertial-Only SLAM: A Fast Pedestrian SLAM Method for Structured Buildings Based on Foot-Mounted IMU

Foot-mounted inertial odometry has facilitated the development of indoor pedestrian positioning. However, achieving long-term accurate pedestrian positioning using inertial sensors remains a significant challenge due to the impact of drift errors. Inertial-only simultaneous localization and mapping (IO-SLAM), which exclusively uses the inertial measurement unit (IMU), effectively mitigates drift in structured buildings and has garnered considerable research interest. However, current IO-SLAM methods suffer from low computational efficiency and sensitivity to randomness, making them less suitable for real-time applications. To overcome these limitations, this paper introduces Fast Inertial-only SLAM (FIO-SLAM), a fast pedestrian SLAM method for structured buildings based on foot-mounted IMU. The proposed method replaces the conventional Rao-Blackwellized particle filter architecture with an optimization-based framework, eliminating randomness introduced by Monte Carlo methods. Additionally, it formulates a constrained optimization SLAM model to accommodate the distinctive features of virtual measurements. Finally, a keyframe optimization strategy is employed to significantly reduce the computational iterations required. Experimental validation in a structured building environment demonstrates the efficacy of FIO-SLAM, which processes 45 minutes of trajectory data (3200 meters) in just 23.3 seconds and achieves an average localization error of 1.67 meters. Compared to state-of-the-art IO-SLAM techniques, FIO-SLAM achieves comparable or better localization accuracy, while requiring only one-tenth of the time cost.