GVIR: A GNSS-Visual-Inertial-Range Sensor Fusion Approach for Seamless Indoor-Outdoor Robot Localization

Global consistent localization for indoor-outdoor robot positioning remains challenging due to complex coordinate system transformations and system switching. To address this, we propose a seamless indoor-outdoor localization method, GVIR, which integrates the Global Navigation Satellite System (GNSS), monocular cameras, Inertial Measurement Unit (IMU), and Range of Ultra-Wideband (UWB) to ensure drift-free and robust global localization. First, a local-global coordinate transformation matrix is estimated to unify the sensor frame across both indoor and outdoor environments. Second, to mitigate GNSS degradation (e.g., obstructions and interference), we propose a robust localization approach based on the Fisher Information Matrix, which evaluates satellite signal quality to counteract GNSS errors. Third, a seamless localization mechanism dynamically adjusts GNSS and UWB weight constraints in indoor-outdoor transition regions to maintain continuous and stable localization. Experimental results show that the system achieves localization accuracy of 0.329m in transition regions.