Multistage Filtering and Intelligent Multi-sensor Fusion Positioning: High-precision Continuous Positioning in Complex Environments

This research proposes a multilevel filtering architecture with intelligent multisensor fusion to achieve high-precision continuous positioning in dynamic indoor-outdoor environments. A three-layer framework integrates Light Detection and Ranging (LiDAR), dual Inertial Measurement Units (IMUs) (built-in and external), and Global Navigation Satellite System (GNSS): primary filters preprocess sensor data via statistical/radius filtering; secondary filters fuse IMUs through adaptive weighting (33.33% error reduction vs. single IMU, max error 0.06 m); tertiary filters deploy an improved extended Kalman filter (EKF) with genetic algorithm-optimized noise matrices. Environmental transitions are managed via a LiDAR-IMU geometric model and fuzzy logic. Experiments on an FR-07 platform show significant accuracy gains: in urban intersections, the improved EKF reduces RMSE to 0.36 m (53.23% lower than traditional EKF) and maximum error to 0.94 m. During GNSS signal obstruction, it maintains 0.23 m RMSE (82.44% lower than traditional EKF) and 0.06 variance, while in high-noise transition environments achieves 2.05 m RMSE (61.54% reduction). The system enables seamless transitions, reducing transitional RMSE by 60.85% at building exits. This architecture enhances robustness in complex measurement scenarios, supporting autonomous navigation.