FusionPortableV2: A Unified Multi-Sensor Dataset for Generalized SLAM Across Diverse Platforms and Scalable Environments

TL;DR

FusionPortableV2, a multi-sensor SLAM dataset featuring sensor diversity, varied motion patterns, and a wide range of environmental scenarios, is presented, demonstrating the dataset’s broad application beyond traditional SLAM tasks by investigating its potential for monocular depth estimation.

Abstract

Simultaneous Localization and Mapping (SLAM) technology has been widely applied in various robotic scenarios, from rescue operations to autonomous driving. However, the generalization of SLAM algorithms remains a significant challenge, as current datasets often lack scalability in terms of platforms and environments. To address this limitation, we present FusionPortableV2, a multi-sensor SLAM dataset featuring sensor diversity, varied motion patterns, and a wide range of environmental scenarios. Our dataset comprises $27$ sequences, spanning over $2.5$ hours and collected from four distinct platforms: a handheld suite, a legged robots, a unmanned ground vehicle (UGV), and a vehicle. These sequences cover diverse settings, including buildings, campuses, and urban areas, with a total length of $38.7km$. Additionally, the dataset includes ground-truth (GT) trajectories and RGB point cloud maps covering approximately $0.3km^2$. To validate the utility of our dataset in advancing SLAM research, we assess several state-of-the-art (SOTA) SLAM algorithms. Furthermore, we demonstrate the dataset's broad application beyond traditional SLAM tasks by investigating its potential for monocular depth estimation. The complete dataset, including sensor data, GT, and calibration details, is accessible at https://fusionportable.github.io/dataset/fusionportable_v2.

Figures (17)

• Figure 1: CAD model of the sensor rig where axes are marked: red: $X$, green: $Y$, blue: $Z$. It visualizes the position of each component of the handheld multi-senosr suite.
• Figure 2: (a) Illustration of data collection which shows the data flow and synchronization processes. The red arrow indicate PPS signals for synchronization, green arrows show UTC time synchronization, and blue arrows represent sensor triggering signals, and black arrows depict the flow of raw data. (b) The timing diagram for triggerable (our case) and non-triggerable sensors, illustrating the unknown time offset caused by the delay in starting data capture, the duration of data capture, and the time required for data transmission from the sensor to the PC. Our synchronization solution can reduce the time delay (i.e., $t_{offset} - t_{delay}$) but cannot address other factors, which require online time calibration algorithms.
• Figure 3: Layouts of the platform-specific sensor setup, including different coordinate systems and their relative translation. More detailed and accurate dimensional data are provided in our calibration files.
• Figure 4: Platform-Specific Data Samples: (a) The handheld multi-sensor rig across various environments, (b) the legged robot, (c) the low-speed UGV, (d) the high-speed vehicle, and (e) the GT generation device. The depicted scenes highlight the FusionPortableV2 dataset's comprehensive coverage across a spectrum of platforms and environmental conditions.
• Figure 5: Sensor placement for the IMU-Prism calibration. Reflective balls for motion capture cameras (MCC) and the prism are marked in red and blue, respectively. We use MCC's measurements to infer high-rate motion of the prism.