This is a ROS2 wrapper around the Basalt VIO library.

Note that the wrapper is split into two parts: frontend and backend. The frontend performs the image feature extraction, the backend runs the graph optimizer etc. The front end publishes messages that contain the extracted features. These messages can be recorded in a rosbag and played back, such that there is no need to store the images to test the backend.

For efficiency reasons, front- and backend are fused together into a single exectuable as well, avoiding the overhead of serializing and de-serializing ROS messages from the Basalt data structures.

Only tested on Ubuntu 18.04LTS with ROS2 Eloquent

This package depends on the Basalt VIO package, which must be installed first by following the instructions for the basalt_ros repository. Failing to do so will give you an error about a missing basaltConfig.cmake. If you don't like the ROS1 packages that come with the basalt_ros repo remove them afterwards, then build by doing this at the root of your workspace:

cd path_to_top_of_your_workspace
colcon build --symlink-install --cmake-args	-DCMAKE_BUILD_TYPE=Release

Overlay your workspace:

cd path_to_top_of_your_workspace
. install/setup.bash   # activate this workspace

Launch the unified front/backend as a node:

ros2 launch basalt_ros2 basalt_vio.launch.py

Now you need supply data from the realsense T265.

To get the calibration file, do this:

ros2 run basalt_ros2 get_calibration  -o my_calib_file.json

If you want to overwrite the imu noise with your own idea of covariance:

ros2 run basalt_ros2 get_calibration  -o my_calib_file.json -a 0.5