Abstract: [Purposes]To address the high computational resource consumption of traditional map-based localization methods in large-scale complex environments,which makes efficient deployment on embedded platforms challenging,this paper proposes a lightweight autonomous localization method integrating LiDAR and inertial navigation.[Methods]The method is built upon sparse map construction,optimizing the scan matching process by extracting local planar features to reduce computational overhead.A time-interval-based map matching strategy combined with a temporal threshold correction mechanism is adopted to avoid per-frame global matching,effectively balancing the computational load.Multi-source sensor data and map matching information are further fused,and a strongly convergent nonlinear geometric observer is introduced to optimize pose estimation.[Findings]Experimental results demonstrate that the proposed method outperforms existing state-of-the-art algorithms in overall performance,with its engineering practicality validated through field inspection deployments at Shengli petrochemical general plant.[Conclusions]The method achieves high-accuracy and robust autonomous localization on resource-constrained embedded platforms,offering a viable technical solution for reliable navigation of unmanned systems in large-scale complex environments．

Key words: inspection robots, LiDAR-based localization, path planning, autonomous localization and navigation