ros2_quad_sim_python

ROS2 packages to simulate (and control) a quadcopter. These packages were created to be used with a flying sensor in CARLA. The flying sensor is a bodiless actor that only has sensors and it will go through the scenario without collisions. The flying_sensor definition can be found here.

How to use:

CARLA Simulator Server

Start the CARLA simulator server in headless mode (https://github.com/ricardodeazambuja/carla-simulator-python):

$ docker run --rm -it \
--name carla-container \
--hostname=carla-container \
--user carla \
-p 2000-2002:2000-2002 \
--gpus 0 \
ricardodeazambuja/carlasim:0.9.13_headless ./launch_headless.sh

ROS2 and the necessary packages

All the instructions can be found in the Dockerfile:

• https://github.com/ricardodeazambuja/ros2_quad_sim_python/blob/main/docker/Dockerfile

Get the script to launch ROS2

• https://github.com/ricardodeazambuja/ros2-playground/blob/main/launch_ros2_desktop.sh

Don't forget to turn the script executable to make life easier ($ chmod +x launch_ros2_desktop.sh).

Launch the container

If you don't want to create your own docker image, you can spawn containers using this command:

$ ./launch_ros2_desktop.sh -g --image ricardodeazambuja/ros2_quad_sim_python

It will mount the current directory inside the container and try to source install/setup.bash.

You can launch as many containers as you need. It's easier than using docker exec for most situations. Inside one of the containers, launch everything:

$ ros2 launch ros2_quad_sim_python launch_everything.launch.py

The launchfile above exposes just a few parameters. You can verify using:

$ ros2 launch ros2_quad_sim_python launch_everything.launch.py --show-args

If you still want to use the same container to save some memory, you need to take note of the containers name (it will be ros2- plus five random characters). Here is how to check the topics without launching a new container (the container's name in this case is ros2-bb93a782c4):

$ docker exec -t ros2-bb93a782c4 bash -i -c "ros2 topic list"

Docker exec is an oneway thing, so I don't recommend using it with anything that needs to catch the ctrl+c (SIGINT) to gracefully stop (like carla_spawn_objects.launch.py needs to remove the actors it created from the CARLA server), it will not work.

Another useful command to use (considering your container's name is ros2-bb93a782c4) is:

$ docker exec -t ros2-bb93a782c4 bash -i -c "ros2 run plotjuggler plotjuggler"

If you just want to start a new terminal using the same container (useful when making changes to that container only during debugging):

$ docker exec -it ros2-bb93a782c4 bash

These are the commands to launch things individually:

$ ros2 launch carla_ros_bridge carla_ros_bridge.launch.py passive:=False town:=Town01
$ ros2 launch carla_spawn_objects carla_spawn_objects.launch.py objects_definition_file:=src/ros-bridge/carla_spawn_objects/config/flying_sensor.json
$ ros2 run rviz2 rviz2 --ros-args -d ~/carla-ros/install/ros2_quad_sim_python/share/ros2_quad_sim_python/cfg/rviz_flying_sensor.rviz

and

$ ros2 run ros2_quad_sim_python quad --ros-args -p init_pose:=[0,0,2,0,0,0] -p Px:=2 -p Py:=2 -p Pz:=1

or

$ ros2 run ros2_quad_sim_python quadsim --ros-args -p init_pose:=[0,0,2,0,0,0]
$ ros2 run ros2_quad_sim_python quadctrl --ros-args -p Px:=2 -p Py:=2 -p Pz:=1

Publish a new setpoint by using $ros2 topic pub /quadctrl/flying_sensor/ctrl_sp quad_sim_python_msgs/msg/QuadControlSetPoint "h and hitting the tab twice so it will fill the rest of the message. Don't forget to remove the - the autocomplete insists adding at the end.
Or publish a new twist setpoint by using ros2 topic pub /quadctrl/flying_sensor/ctrl_twist_sp geometry_msgs/msg/Twist "l and following the same hints from above.

Another option is to control the simulated quadcopter using the keyboard with the teleop_twist_keyboard package (where flying_sensor is the default name):

$ ros2 run teleop_twist_keyboard teleop_twist_keyboard --ros-args -r teleop_twist_keyboard:cmd_vel:=/quadctrl/flying_sensor/ctrl_twist_sp

Time synchronisation

Now the quadsim node is calculating the difference between the machine time and the time CARLA ros-bridge publishes. This time difference is available as the parameter carla_time_diff_ns. The node quadctrl reads this parameter from quadsim. All messages are synchronised with CARLA ros-bridge, but the assumption is the hosts have their system times synchronised.

Training Episode Generator

• spawn_episode.py: randomly choose a spawn point from the map and populate the area around with vehicles and walkers.

Modifying the weather inside CARLA

Below is an example of how to modify the weather:

$ ros2 topic pub /carla/weather_control carla_msgs/msg/CarlaWeatherParameters "cloudiness: 30.0 
precipitation: 0.0
precipitation_deposits: 0.0
wind_intensity: 0.0
fog_density: 0.0
fog_distance: 0.0
wetness: 0.0
sun_azimuth_angle: 15.0
sun_altitude_angle: 15.0" --once

Making changes

Don't forget to run colcon build --symlink-install if you change anything that is not just a Python script.

TODO

• Add an example publishing setpoints to quadctrl, reading the current state, etc.
• Add an example using the wind.
• Create a message for the potential field and update ros_quad_ctrl to subscribe to a topic receiving that message.
• Make the topic where the simulator publishes the poses a parameter so it can be easily changed allowing it to be used with other things than CARLA.
• Consider a better way to synchronise time between the ROS2 nodes.