EU Long-term Dataset with Multiple Sensors for Autonomous Driving

Zhi Yan, Li Sun, Tomas Krajnik, and Yassine Ruichek

Issues

• Radar data: 13
• Vehicle position ground-truth provided by GPS-RTK: 1, 7

Dataset

https://epan-utbm.github.io/utbm_robocar_dataset/

Baselines

We forked the implementation of the following state-of-the-art methods and experimented (with minor changes) with our dataset:

• Pose Estimation: https://github.com/tu-darmstadt-ros-pkg/hector_localization
• Lidar odometry: https://github.com/laboshinl/loam_velodyne
• Lidar odometry: https://github.com/RobustFieldAutonomyLab/LeGO-LOAM
• To be continued ...

All users are more than welcome to commit their results.

Ground-truth trajectories recorded by GPS/RTK

hector_pose_estimation (pose estimation)

How to play

roslaunch utbm_pose_estimation.launch bag:=path_to_your_rosbag

utbm_pose_estimation.launch is here.

Evaluation

First of all, you should have something like this:https://yzrobot.github.io/

Then, /fix/pose(geometry_msgs/PoseStamped) ihttps://yzrobot.github.io/s the estimated robot pose (6DOF) based on the GPS data.

loam_velodyne (lidar odometry)

How to play

roslaunch loam_velodyne loam_velodyne_utbm.launch bag:=path_to_your_rosbag

loam_velodyne_utbm.launch is here.

Evaluation

First of all, you should have something like this:

Then, /aft_mapped_to_init(nav_msgs/Odometry) is the output lidar odometry that needs to be evaluated.

Single Velodyne HDL-32E (left)

LeGO-LOAM (lidar odometry)

1. Minor changes w.r.t. LeGO-LOAM (Issue #9, thanks to robot54)

How to play

roslaunch lego_loam lego_loam_utbm.launch bag:=path_to_your_rosbag

lego_loam_utbm.launch is here.

Evaluation

First of all, you should have something like this:

Then, /aft_mapped_to_init(nav_msgs/Odometry) is the output lidar odometry that needs to be evaluated.

Single Velodyne HDL-32E (left)