The Worldwide Journal of Robotics Analysis, Forward of Print.
Simultaneous Localization and Mapping (SLAM) has been broadly utilized in varied robotic missions, from rescue operations to autonomous driving. Nevertheless, the generalization of SLAM algorithms stays a major problem, as present datasets usually lack scalability when it comes to platforms and environments. To handle this limitation, we current FusionPortableV2, a multi-sensor SLAM dataset that includes sensor range, diversified movement patterns, and a variety of environmental eventualities. Our dataset includes 27 sequences, spanning over 2.5 hours and picked up from 4 distinct platforms: a handheld suite, a legged robotic, an unmanned floor automobile (UGV), and a automobile. These sequences cowl various settings, together with buildings, campuses, and concrete areas, with a complete size of 38.7 km. Moreover, the dataset contains floor fact (GT) trajectories and RGB level cloud maps masking roughly 0.3 km2. To validate the utility of our dataset in advancing SLAM analysis, we assess a number of state-of-the-art (SOTA) SLAM algorithms. Moreover, we reveal the dataset’s broad utility past conventional SLAM duties by investigating its potential for monocular depth estimation. The entire dataset, together with sensor information, GT, and calibration particulars, is accessible at https://fusionportable.github.io/dataset/fusionportable_v2.
发布者：Hexiang Wei，转转请注明出处：https://robotalks.cn/fusionportablev2-a-unified-multi-sensor-dataset-for-generalized-slam-across-diverse-platforms-and-scalable-environments/
