The following content is from the daily update content of the robot SLAM Learning Circle Knowledge Planet of Primary Six
#thesis# arxiv | Carnegie Mellon University Releases SubT-MRS: An Underground, Multi-Robot, Multispectral, and Multi-Degenerate Robust SLAM Dataset
【SubT-MRS: A Subterranean, Multi-Robot, Multi-Spectral and Multi-Degraded Dataset for Robust SLAM】
Link to the article: http://arxiv.org/abs/2307.07607
In recent years, significant progress has been made in the field of Simultaneous Localization and Mapping (SLAM). However, the current state-of-the-art solutions still face limited accuracy and robustness in practical applications. One of the main reasons for this is the lack of datasets that adequately capture the conditions faced by robots in the field. To solve this problem, we propose SubT-MRS, an extremely challenging real-world dataset designed to push the limits of SLAM and perception algorithms.
SubT - MRS is a multimodal, multi-robot dataset collected primarily from subsurface environments with multiple degraded conditions, including unstructured corridors, varying lighting conditions, and perceptual occlusions such as smoke and dust. In addition, the dataset encapsulates information from time-synchronized sensors of different ranges, including lidar, vision cameras, thermal cameras, and IMUs captured using different vehicle movements such as airborne, leg, and wheeled, to support the study of sensor fusion, which is essential for achieving accurate and robust robot perception in complex environments.
To evaluate the accuracy of the SLAM system, we also provide a dense 3D model with sub-centimeter accuracy, as well as accurate 6DoF ground truth. Our benchmarking methodology includes several state-of-the-art methodologies to demonstrate the challenges introduced by our datasets, especially in the context of multiple degraded environments.
The above content is from the daily update content of the robot SLAM Learning Circle Knowledge Planet of Primary Six
Recommended Reading:
The World's First Open Source Autonomous Driving Integration Framework Autoware!
Deep learning feature extraction matches open-source algorithms SuperPoint and SuperGlue