A Multi-Scenario Dataset for Long-Term Indoor Localization and Pedestrian Behavior Analysis in Dynamic Environments

Human activity and structural modifications continuously alter shared indoor spaces, leading to challenging conditions for reliable localization and motion understanding. To investigate and analyze the impact of such dynamics on long-term indoor localization, we present a multi-scenario dataset designed under controlled levels of occlusion and environmental change. The data were collected in a university entrance hall configured to simulate a conference environment, with movable poster walls and natural pedestrian activity around them. A movable LiDAR platform was used to collect data within the environment, while four synchronized overhead AI cameras captured multi-view pedestrian motion. The image data from the cameras are synchronized with the LiDAR point clouds, enabling joint analysis of pedestrian behavior in both 2D and 3D domains. Three scenarios, named extreme occluded, semi occluded, and free space, represent increasing levels of structural modification and visibility loss. High-precision ground truth was established using total station tracking. The dataset enables systematic research on localization performance under evolving indoor conditions and supports the analysis of pedestrian behavior and human–robot interaction in shared spaces.