AirMuseum: a heterogeneous multi-robot dataset for stereo-visual and inertial Simultaneous Localization And Mapping

Abstract

This paper introduces a new dataset dedicated to multi-robot stereo-visual and inertial Simultaneous Localization And Mapping (SLAM). This dataset consists in five indoor multi-robot scenarios acquired with ground and aerial robots in a former Air Museum at ONERA Meudon, France. Those scenarios were designed to exhibit some specific opportunities and challenges associated to collaborative SLAM. Each scenario includes synchronized sequences between multiple robots with stereo images and inertial measurements. They also exhibit explicit direct interactions between robots through the detection of mounted AprilTag markers [1]. Ground-truth trajectories for each robot were computed using Structure-from-Motion algorithms and constrained with the detection of fixed AprilTag markers placed as beacons on the experimental area. Those scenarios have been benchmarked on state-of-the-art monocular, stereo and visual-inertial SLAM algorithms to provide a baseline of the single-robot performances to be enhanced in collaborative frameworks.

Towards Collaborative Simultaneous Localization and Mapping: a Survey of the Current Research Landscape

“…Moreover, only a few datasets dedicated to C-SLAM exist. (Leung et al, 2011a) consists of 9 monocular camera subdatasets and (Dubois et al, 2020a) is dedicated to stereo-inertial C-SLAM. Therefore, the common approach to evaluate C-SLAM solutions is to split single robot SLAM datasets into multiple parts and to associate each one to a robot.…”

Sharing visual-inertial data for collaborative decentralized simultaneous localization and mapping

“…However, the EuRoC sequences were primarily designed for singlerobot SLAM, and consequently, trajectories are already self-sufficient in terms of loop closures. To properly stage the specific challenges of online collaborative SLAM, we designed the AirMuseum dataset [2]. It consists in five heterogeneous multi-robot scenarios with aerial and terrestrial sequences whose properties are displayed in Table 1b.…”

“…They were designed to be agnostic to the SLAM Front-End as long as it outputs a visual-inertial factor graph. We evaluate them on multi-robot scenarios built from the EuRoC dataset [1] and our custom AirMuseum dataset [2].…”

“…We also propose a common collaborative SLAM architecture to organize the computation, exchange and integration of such packets. We evaluated those methods on the EuRoC [1] dataset and on our custom dataset AirMuseum [2]. Experiments showed that the proposed methods allow the agents to build, exchange and integrate consistent visual-inertial packets, and improve their trajectory estimation accuracy up to several centimeters.…”

“…We advocate it allows a fairer comparison between the mono and multi-robot scenarios which thus share the very same underlying SLAM algorithm. The performances of the actual Vins-Mono algorithms on the EuRoC and AirMuseum sequences are respectively reported in[10] and[2].…”