26 May 2026 in IEEE Transactions on Instrumentation and Measurement
Classical visual-inertial simultaneous localization and mapping (VI-SLAM) systems provide accurate relative motion estimation but are plagued by accumulated drift in long-term GNSS-denied environments where loop closures are unavailable. To address this fundamental challenge, we propose a novel framework that leverages “Information of Opportunity” (IOP)—random and heterogeneous diverse external signals including terrestrial radio signals, Low Earth Orbit (LEO) satellite observations or map data to achieve drift-less geo-localization. Our system integrates three core components: (1) an enhanced visual-inertial front-end with information-theoretic feature selection( FS) that adaptively selects informative visual features for robust and efficient tracking; (2) a versatile IOP processing pipeline that acquires and aligns absolute positioning cues (e.g., from LEO pseudoranges or image-to-map matching) into a unified global reference frame for enforcing global drift-free constraints; and (3) a tightly-coupled optimization back-end that jointly infers IOP-derived global constraints with local visual-inertial measurements within a unified factor graph, enabling continuous drift-less localization. Extensive evaluations on the KITTI Odometry dataset, GVINS dataset and real-world urban driving experiments, conducted with stereo-infrared cameras, IMU, and Iridium LEO receiver, demonstrate that our method effectively suppresses cumulative drift and achieves globally driftless localization.
1 January 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing