GMU Cartographer setup to read Kitti Database.
Kitti Raw Datafiles available at: http://www.cvlibs.net/datasets/kitti/raw_data.php
Kitti2bag available at: https://github.com/tomas789/kitti2bag
I made minor modifications to Kitti2bag, and its source is included in scripts directory.
Top of my head things.
#1 - I removed all the stuff it was packaging into the bag file except the IMU & Velodyne Points. #2 - I renamed the frame_ids, and topics to my liking. (they match the configurations) #3 - I removed the tf topic it was publishing and instead created a robot URDF file.
Note: the layout of the URDF file should be based off: http://www.cvlibs.net/datasets/kitti/setup.php
Both Lidar, and IMU. X is direction travelled in car, Y points to driver side of car, and Z points straight up.
Lidar is in center of car, IMU is 0.32m to the driver side (left) of car in Y direction. IMU is also positioned behind Lidar 0.81m (in negative X direction) Lidar is at height (z) 1.73m, while IMU is at height 0.93m
I am using the the imu as the frame_id for slam.
For convience of google, I have uploaded one of the bag files.
https://s3.amazonaws.com/kittigmu/kitti_2011_10_03_drive_0027_synced.bag
There should be 5 loop closures, and if you play the bag through to the end, it should end up roughly where it started. I have provided a PNG file for reference.
Any help or suggestions would be apprechated.
Examples:
roslaunch gmu kitti.launch bag_filenames:=/home/user/kitti.bag
roslaunch gmu kitti_write.launch bag_filenames:=/home/user/kitti.bag pose_graph_filename:=/home/user/kitti.bag.pbstream