SensNav: A Multisensor Vehicle Navigation Dataset for Campus and Urban Environments With a Long Tunnel

The rapid development of the mechanical industry, automated manufacturing, and artificial intelligence accelerates the evolution of autonomous driving systems. Decision-making algorithms for autonomous vehicles require extensive environmental information captured by multiple sensors to execute vehicle control commands in complex and dynamic environments. Consequently, multi-sensor datasets serve as valuable references and training resources for autonomous driving research. However, existing datasets suffer from significant limitations in long tunnel coverage, diverse campus environments, raw global navigation satellite system (GNSS) data, sky-view fisheye camera imagery, and data lightweighting. To address these gaps, we developed a customized multi-sensor vehicle-mounted platform and established the SensNav dataset. The SensNav dataset is characterized by: 1) a long trajectory of 18.62 km, including a 5.2 km tunnel; 2) various campus scenes under both day and night conditions, featuring open-sky areas, tree-lined avenues, and regions shaded by buildings; 3) raw GNSS measurements from a 10 Hz receiver and a 20 Hz receiver; and 4) a sky-view fisheye camera. Specifically, the campus dataset contains three types of inertial measurement units (IMUs), a forward-facing RGB camera, a fisheye camera, and two GNSS receivers, while the urban dataset comprises two GNSS receivers and two IMUs. A high-precision integrated GNSS/inertial navigation system (INS) was employed, and post-processing techniques were applied to obtain accurate ground truth trajectories. This dataset provides critical experimental support for emerging research on GNSS positioning enhancement, multi-sensor fusion algorithms, and navigation system evaluation under complex operational conditions. The SensNav dataset is publicly available at https: //github.com/dihe2012/SensNav-Dataset.git.