Loop Closure Detection
Overview
Loop closure detection asks whether a robot has returned to a mapped place, enabling SLAM to correct drift and keep trajectories and maps globally consistent. Systems typically retrieve candidates with visual or LiDAR place-recognition descriptors, then reject perceptual aliases using geometric registration, sequence or semantic consistency, and add accepted constraints to pose graphs. Research spans visual-inertial benchmarks and challenging settings such as repetitive LiDAR routes, dynamic thermal scenes, forests, radar, and sonar, with learned features, topology, and multimodal sensing extending beyond conventional appearance matching. The central tension is accepting enough true revisits to limit drift while rejecting false matches under tight compute budgets and changing viewpoints, environments, and sensors; extra semantic or spectral cues do not always improve discrimination.
Latest papers
22 papers3 papers in the last four weeks, against 1 the four weeks before. 0.0% of all new papers.