prototype vehicle

Note in either vehicle case, ROS master must be the PC. Thus before any of the below, launch roscore on the PC.

Setting up symlinks/udev rules

After cloning this repo, copy the udev rules in the 'config' folder to the udev rules directory with

sudo cp config/99-r88.rules /etc/udev/rules.d/

Then load the rules with

sudo udevadm control --reload-rules && udevadm trigger

Unplug and replug any USB devices for the changes to immediately take effect (or just reboot the CPU).

RPi Drone

First ssh into the drone:

ssh [email protected]

Launch MAVROS on the drone. Note that the ROS Master has already been set to the host PC. This command works when Pixhawk connected to Pi by USB:

roslaunch mavros apm.launch

If MAPIR NIR camera is plugged into Pi, start MAPIR image publishing with:

roslaunch vehicle_launch mapir_nir.launch

On PC

Launch all offboard perception with (remove last arg once Sequoia camera integrated):

roslaunch vehicle_launch prototype_offboard.launch do_multispectral_viz:=false

Nano Drone

First ssh into the drone (showing static IP for NETGEAR88-5G):

ssh [email protected]

Launch all nodes with (default is ZED + Mapir, include last arg if also Attollo):

roslaunch vehicle_launch prototype.launch use_attollo:=true

On PC

Launch vizualization with:

roslaunch vehicle_launch viz_host_machine.launch

Cameras

At the moment, camera device IDs often change on start-up. Until we can automatically detect and assign IDs, check for camera IDs with:

v4l2-ctl --list-devices

Then update the device_id param in the respective launch files for Attollo and Mapir.

post-processing

To run postflight visualization

roscore
rosbag play *.bag --pause --clock
roslaunch vehicle_launch viz_host_machine.launch

To run post-processing:

roscore
rosbag play *.bag --pause --clock /tf:=/tftrash /cloud_registered:=/cloudtrash /costmap_node/costmap/costmap:=/costmaptrash /tf_static:=/statictrash
roslaunch vehicle_launch decco.launch offline:=true slam_type:=...

Note in post-processing, slam_type can be changed but must match the incoming pointcloud type (e.g., slam_type 0 expects Livox data as sensor_msgs/Pointcloud2, and slam_type 1,2 expects custom_livox_msg). TODO: update decco.launch and move any unnecessary nodes to a group launched only in offline mode.

Texas datasets

For the TX field tests, we had to separately run Attollo and MAVROS, so initialization data is in a separate bag from the flight bag. Combination is manual. First, run the above post-processing command, and wait until map yaw, utm_x, and utm_y print in terminal. Then in task_manager/launch/task_manager.launch, update the values accordingly and set explicit_global_params to true in that same file. Then run the above post-processing command but with the inflight bag.

recording

TODO update this section and delete deprecated launch files.

To record only the inputs (i.e., for a bag to be used in development), run:

roslaunch vehicle_launch rosbag_record_inputs.launch record_prefix:=<your dir>

^This should generally be preferred compared to below. Recording all topics risks overrunning the buffer significantly and losing all data.

To record all required variables, input or output (i.e., for a bag from flight testing for later analysis), run:

roslaunch vehicle_launch rosbag_record_all_io.launch record_prefix:=<your dir>

After flight, if the rosbag record terminal shows warning messages about the buffer overrun, stop the prototype ROS nodes before the bag, and give some time (at least 60sec) before stopping the rosbag node. Verify bags by:

cd ~/dev/bags
rosbag info <yourbag>.bag

Then compress all valid bags (bags with any topics shown) with rosbag compress *.bag after moving all valid bags to a subfolder. This will significantly speed up file transfer.

If jpegs are needed, run

roslaunch vehicle_launch jpeg_from_bag.launch bag:=<your bag>,

then mv ~/.ros/frame*.jpg <desired dir>.

camera calibration

Requires a checkerboard. Generates the camera info yaml for undistorting images and 3D projection.

Example uses MAPIR camera. First, launch the camera ROS driver:

roslaunch vehicle_launch mapir_nir.launch

Then, run the camera calibrator:

rosrun camera_calibration cameracalibrator.py --size 7x7 --square 0.05715 image:=/mapir_cam/image_raw camera:=/mapir_cam

A more detailed tutorial for the calibrator can be found here.

GoPro

When connected by USB, the GoPro is not available for ROS until the following is run:

sudo gopro webcam -p enx -n -a

after following the instructions at this repo. Note GoPro still needs a camera calibration yaml using the standard chessboard process.