EUFS Autonomous Datasets

This repo has moved to gitlab - https://gitlab.com/eufs/datasets

A collection of datasets gathered by the EUFS team whilist developing an autonomous formula student car.

All datasets come in a rosbag format and can be used directly in ROS. Otherwise, raw data can be extracted using bag_tools.

Repository also includes custom ROS messages featured in the datasets.

Contents

1. Sanderson car park
2. FS-AI 2018 Dataset

Sanderson car park

Collected in a car park using a testbed with sensors attached at the appropriate positions.

• single lap - download

• double lap - download

Topics

Topic	Type	Freq	Source	Notes
/imu/data	sensor_msgs/Imu	10Hz	X-Sens MTi-G	6 DoF IMU Data from X-Sens MTi-G INS with embedded Kalman Filter. Includes orientation data.
/imu/mag	sensor_msgs/MagneticField	10Hz	X-Sens MTi-G
/fix	sensor_msgs/NavSatFix	10Hz	X-Sens MTi-G	Unreliable as far as our tests go.
/velodyne_points	sensor_msgs/PointCloud2	10Hz	Velodyne VLP-16	Full 360 degree data. Testbed must be filtered out.
/left/image_rect_color/compressed	sensor_msgs/CompressedImage	30Hz	ZED Camera	672 x 376 resolution
/right/image_rect_color/compressed	sensor_msgs/CompressedImage	30Hz	ZED Camera	672 x 376 resolution
/odom	nav_msgs/Odometry	30Hz	ZED Camera	Visual odometry. Very good!!
voxel_frid/output	sensor_msgs/PointCloud2	30Hz	ZED Camera	Slightly filtered 3D pointcloud from the stereo camera.
/tf	tf2_msgs/TFMessage	30Hz	-	Contains odom->base_link published by the ZED camera
/tf_static	tf2_msgs/TFMessage	once	-	Transforms for all sensors on the car

Transforms [x, y, z, r, p, y]

base_frame is center of gravity aligned to the lowest position of the car.

• IMU: [0, 0, 0, 0, 0, 0]
• GPS: [0, 0, 0.82, 0, 0, 0]
• VLP-16: [1.541, 0, 0, 0, 1.5*PI/180.0, 0]
• ZED camera: [0.2, 0, 0.471, 0, 0, 0]

Also embedded in /tf_static

Notes

• IMU data is a bit unusable due to the low output frequency.
• Test ground was not flat.

FS-AI 2018 Dataset

Collected at FSUK 2018 using the ADS-DV vehicle provided by the IMechE. This should be a representative of sensory input from a mule vehicle in the DDT class of future FSUK competitions.

download

Topics

Topic	Type	Freq	Source	Notes
/imu/data	sensor_msgs/Imu	167Hz	X-Sens MTi-G	6 DoF IMU Data from X-Sens MTi-G INS with embedded Kalman Filter. Includes orientation data.
/imu/mag	sensor_msgs/MagneticField	167Hz	X-Sens MTi-G
/fix	sensor_msgs/NavSatFix	167Hz	X-Sens MTi-G	Unreliable as far as our tests go. In reality changes at 10Hz
/velodyne_points	sensor_msgs/PointCloud2	10Hz	Velodyne VLP-16	Full 360 degree data. Wheels of the car must be filtered out.
/velocity	geometry_msgs/TwistStamped	167Hz	X-Sens MTi-G	Velicty from Kalman Filter of the GPS. In reality changes at 10Hz.
/left/image_rect_color/compressed	sensor_msgs/CompressedImage	32Hz	ZED Camera	672 x 376 resolution
/right/image_rect_color/compressed	sensor_msgs/CompressedImage	32Hz	ZED Camera	672 x 376 resolution
/odom	nav_msgs/Odometry	32Hz	ZED Camera	Visual odometry. Useless due to bad positioning of the camera.
/tf_static	tf2_msgs/TFMessage	once	-	Transforms for all sensors on the car
/ros_can/state_str	std_msgs/String	100Hz	Car ECU	High-level state of the car
/ros_can/wheel_speeds	eufsa_msgs/wheelSpeeds	100Hz	Car ECU	Wheel RPM and steering feedback from the car. Custom message is simply a header and 5x float32; can be found in repositoty.

Transforms

embedded into /tf_static

Notes

• Full data was not collected due to lack of time and equipment.
• Due to issues with the car, the steering feedback, in reality, is only 1Hz. Use with caution.
• High-frame-rate camera data was not recorded due to insufficient bandwidth of the hardware on the day. In reality, it runs at 100Hz.
• Visual odometry was rendered useless because 1/3 of the camera frame was the car itself. Take note.
• GPS data has not been tested.
• In our tests, we couldn't get a good wheel odometry to work due to the bad steering feedback. It should be fixed in the future and should output at 100Hz as well.