Standalone ROS 2 Jazzy sensor and Nav2 autonomy bringup for the DCIST Clearpath Jackal.
This repository is packaged as a sub-module for the main jackal autonomy image repository
Run sensors and navigation in separate terminals so either side can be restarted and inspected independently:
ros2 launch jackal_nav2 jackal_sensors.launch.py
ros2 launch jackal_nav2 jackal_navigation.launch.py
The sensor launch starts Ouster, ZED, and DLIO by default. The navigation launch starts GroundGrid, the map/odometry static transforms, the tested Nav2 server set, and the physical-velocity-to-Joy bridge. The Jackal platform driver that consumes /joy must already be running.
The sensor launch also publishes os_lidar -> zed_camera_link when ZED is enabled.
config/zed_mount.yaml defaults to mode: calibration: it composes the retained
LiDAR-to-left-optical calibration with the inverse of the camera root-to-optical
transform from the ZED description. DLIO owns odom -> base_link -> os_lidar;
the ZED description owns all internal camera frames. This keeps TF consistent
with fusion without changing the calibration, DLIO or scan-time projection code.
The profile specifies the calibration file, parent/child frames, ZED description
package/path, camera name/model and custom baseline. These must match the running
camera. The historical x_m, y_m, z_m, roll_rad, pitch_rad, yaw_rad
values apply only to explicit mode: manual; set its parent accordingly. They are
not used as a fallback when calibration loading fails. The current DLIO base
origin convention is preserved, not physically recalibrated.
To publish only the mounting edge:
ros2 launch jackal_nav2 zed_mount.launch.py
ros2 run tf2_ros tf2_echo base_link zed_left_camera_optical_frameThe complete lookup also requires DLIO and the ZED internal TF publisher. Use one
owner for the mounting edge. If the standalone publisher is already running, use
publish_zed_mount_tf:=false with jackal_sensors.launch.py, or stop it before
restarting sensors. Navigation's publish_zed_static_tf defaults to false; if
explicitly enabled it uses the same profile. All three launch paths accept
zed_mount_config:=/absolute/path/to/config.yaml. Calibration mode forbids
relabeling its LiDAR parent as a body frame. Restart consumers when replacing an
old base-to-camera publisher so their TF caches do not retain the old parent.
Send a waypoint file with:
ros2 run jackal_nav2 goto_nav2 /path/to/waypoints.yaml
The waypoint format contains a frame_id and a waypoints list. Each waypoint accepts x, y, and an optional yaw value.
Record a navigation-focused diagnostic bag with:
ros2 launch jackal_nav2 record_jackal.launch.py profile:=navigation
Available profiles are navigation, lidar, semantic, and full. Use
extra_topics:=/topic/one,/topic/two to append experiment-specific topics.
The surrounding workspace or robot image is responsible for supplying:
- ROS 2 Jazzy and Navigation2, including MPPI, rotation shim, SmacPlanner2D, velocity smoother, and SpatioTemporalVoxelLayer.
- ouster_ros
- zed_wrapper
- direct_lidar_inertial_odometry
- groundgrid
The optional motion_stats node compares the raw Nav2 command, velocity-smoother
output, and measured odometry. Enable it with:
ros2 launch jackal_nav2 jackal_navigation.launch.py \
start_motion_analysis:=true
It updates the current run's artifacts every 30 seconds and once more on clean
shutdown. Each run is stored under plots/run_<timestamp>/ in the source package
(or ~/.ros/jackal_nav2/plots when no source workspace can be found):