Particle-Filter-Based LiDAR Localization for Indoor Parking Lot

Accurate localization of autonomous vehicles in indoor environments is challenging due to the absence of GPS signals, so various studies have explored the use of environmental sensors to address this limitation. In this paper, we propose an indoor localization algorithm that utilizes a 3D LiDAR sensor and a 2D map, supported by improved motion and sensor modeling tailored for indoor parking lots. These environments contain complex conditions in which static noise from parked vehicles and dynamic noise from moving vehicles coexist, requiring a localization method capable of maintaining robustness under high-noise conditions. In this study, vehicle odometry was obtained using LOAM-style scan-to-scan LiDAR odometry, and a particle filter was implemented based on this information. The proposed algorithm was validated using a test vehicle in two indoor parking lots under three different conditions: when the lot was empty, when parked vehicles were present, and when other moving vehicles were present. Experimental results demonstrated that the algorithm achieved an average localization error of approximately 0.09 m across all scenarios, confirming its effectiveness for indoor parking environments.