multisensor-SLAM

Software installation

Software dependencies and installation procedure are specified in the "setup/installation"-folder.

Download KITTI data and generate rosbag

Download only the 00 sequence of the kitti dataset. The script also generates a rosbag from the kitti data.

$ cd setup/download/
$ ./download-kittidata-00-full

Setup

$ git clone https://github.com/tshellum/multisensor-SLAM.git
$ cd multisensor-SLAM
$ git submodule update --init --recursive
$ cd setup/
$ ./uncompress_vocabulary.sh

Usage

Modify the config files to fit the dataset that is to be used. Then run:

$ roscore
$ catkin build
$ roslaunch vo kitti.launch
$ roslaunch backend kitti.launch

or

$ roscore
$ catkin build
$ ROS_NAMESPACE=camera_array rosrun stereo_image_proc stereo_image_proc
$ roslaunch vo ma2.launch
$ roslaunch backend ma2.launch

To visualize the generated point cloud and the motion of the vessel, type:

$ roslaunch cloud_viewer viz.launch

To save the motion of the vessel onto a txt file, type:

$ roslaunch motion2file eval.launch

Then play the rosbag

$ rosbag play /path/to/rosbag

To visualize images use rviz or

$ rosrun image_view stereo_view stereo:=camera_array image:=image_rect

Example: For the 00 sequence of the rosbag it is recommended to play from 3 seconds in becasue the measurement rate is for some reason lower initially.

$ rosbag play -s 3 ~/Videos/kitti/kitti_2011_10_03_drive_0027_sync.bag

Visualization

Plotting the optimization graph

Run the system following the instructions above. Convert the created .dot file to a visual file using the following command.

dot -Tps graph.dot -o graph.ps

Plotting the estimated trajectory compared to GNSS data

This module also computes the ATE and RTE for the odometry vs the ground truth data based on the closest timestamps.

$ rosrun rpg_trajectory_evaluation analyze_trajectory_single.py results/ --recalculate_errors