Replication Data for: VIGS-Fusion:Real-time Inertial Gaussian Splatting SLAM onboard the UAV

VIGS-Fusion Dataset VIGS-Fusion Dataset This dataset contains 8 sequences taken onboard our UAV in ROS2 bag files. The sequences represent various types of movements, including linear, rotational, and combined motions, and are designed to be used for testing and developing visual inertial algorithms on UAV. Topics Included in Each Sequence /Darknav/pose: Provides the ground truth pose of the UAV. /camera/color/camera_info and /camera/depth/camera_info: Provide the camera parameters. /camera/color/image_raw/compressed: The color image compressed. /camera/depth/image_rect_raw: The depth image. /camera/imu: The RealSense IMU (not used). /filtered_imu_xsens: Provides the IMU measurements at 200 Hz. Transformation Between IMU and Camera The transformation between the IMU and the camera is given as follows: # camera_color_optical_frame xsens_imu_link # tx ty tz qx qy qz qw -0.021 -0.073 -0.046 -0.504 0.506 -0.505 0.484 Sequences in the Dataset The dataset consists of the following sequences, each stored in a separate ROS2 bag file: Line: A straight-line motion followed by a return to the starting position. Square-Shape: A square-shaped trajectory with only translations. Rotation-Only: Rotation around the yaw axis only. XYZ-Rotation: Combined translational (X, Y, Z) and rotational motion. XYZ-Rot-obstacle: Combined translational and rotational motion with many obstacles across the scene. XYZ-Rot-Fast: Combined translational and rotational motion at a high speed. Square-Shape-2: A square-shaped trajectory around a box in the center. Line-Rot: Straight line movement, the robot looks around, and then returns to the start.

VIGS-Fusion 数据集（VIGS-Fusion Dataset）
本数据集包含8个由我们的无人机（Unmanned Aerial Vehicle，UAV）采集的ROS2包格式序列。这些序列涵盖直线运动、旋转运动及复合运动等多种运动类型，旨在用于无人机平台上视觉惯性算法的测试与开发。

### 各序列包含的ROS话题
/Darknav/pose：提供无人机的真值位姿。
/camera/color/camera_info 与 /camera/depth/camera_info：提供相机参数。
/camera/color/image_raw/compressed：压缩格式彩色原始图像。
/camera/depth/image_rect_raw：校正后的原始深度图像。
/camera/imu：RealSense 惯性测量单元（Inertial Measurement Unit，IMU），本数据集未使用该设备的数据。
/filtered_imu_xsens：提供200Hz采样频率的滤波xsens IMU测量数据。

### IMU与相机的外参变换
IMU与相机之间的外参变换如下：
# camera_color_optical_frame xsens_imu_link
# tx ty tz qx qy qz qw
-0.021 -0.073 -0.046 -0.504 0.506 -0.505 0.484

### 数据集中的序列
本数据集包含以下8个序列，每个均存储于独立的ROS2包文件中：
1. Line：先沿直线运动后返回起始位置的直线往返运动。
2. Square-Shape：仅包含平动的正方形轨迹。
3. Rotation-Only：仅围绕偏航轴的旋转运动。
4. XYZ-Rotation：包含X、Y、Z轴平动与旋转的复合运动。
5. XYZ-Rot-obstacle：场景中存在大量障碍物的复合平动与旋转运动。
6. XYZ-Rot-Fast：高速复合平动与旋转运动。
7. Square-Shape-2：围绕中心箱体的正方形轨迹运动。
8. Line-Rot：直线运动期间机器人环视周围环境，随后返回起始位置。

搜集汇总

背景与挑战

背景概述

该数据集专为无人机视觉惯性SLAM算法开发设计，包含8个ROS2 bag序列，涵盖线性、旋转和组合运动等多种典型场景。数据集提供多模态传感器数据，包括彩色图像、深度图像、IMU测量（200 Hz）和地面真值位姿，并附有IMU与相机之间的变换参数，适用于算法测试与验证。