isar-turtlebot

ISAR implementation for the Turtlebot3 Waffle Pi.

ISAR - Integration and Supervisory control of Autonomous Robots - is a tool for integrating robot applications into Equinor systems. Through the ISAR API you can send command to a robot to do missions and collect results from the missions.

Running the full ISAR system requires an installation of a robot which satisfies the required interface. isar-turtlebot is an ISAR implementation of the Turtlebot3 Waffle Pi, which enables running the Turtlebot with ISAR through simulations or physically.

Run the simulation using docker

Give Docker containers access to the Nvidia graphics card on the host machine

Pre-requisites

• git
• docker
• docker-compose

NOTE: Docker must NOT be installed using Snap. The Snap version is not compatible with nvidia-docker2. Instead, follow the official documentation from Docker for installation.

To check if your Docker version was installed using snap, run the following command

systemctl list-units --type=service | grep docker

If the result is snap.docker.dockerd.service, the installation has been done using snap and docker must be reinstalled.

Installation nvidia docker

1. Install the newest recommended drivers from Nvidia if not already as described by the following documentation
2. Install nvidia-docker2 using the official documentation

Run simulation

Build the container. This needs to be done once before one can give the container access to the screen.

docker-compose build

Give the docker container access to the screen, this needs to be done each time the computer is restarted.

xhost +local:`docker inspect --format='{{ .Config.Hostname }}' turtle_sim`

Start the simulation

docker-compose up

To build and start the simulation

docker-compose up --build

The simulation world that is used can be set by changing the world variable in the 'entrypoint.sh' file.

Additional settings, such as using Nvidia GPU or gamepad input to the docker container is included via separate .yml-files. To run the docker container with these settings the corresponding docker-compose-setting.yml file must be spesified together with the main docker-compose.yml file. Several setting files can be included.

docker-compose -f docker-compose.yml -f docker-compose-setting.yml up --build

To run the simulation in headless mode set the environment variable HEADLESS=true prior to launching the docker container. Alternatively directly as an environment variable in the docker-compose command:

HEADLESS=true docker-compose up --build

The simulation can then be viewed at webviz with the following link: https://webviz.io/app/?rosbridge-websocket-url=ws://localhost:9090/

Video Stream

A video stream of the front camera is found on the address

http://localhost:5000/stream_viewer?topic=/camera/rgb/image_raw

Adding new models

New models can be added by placing the model that is used by Gazebo into '/models/new_world/' and adding a "new_world.world" file into 'ros_packages/isar_turtlebot/worlds/'. The map that is used by the planner should be placed into '/ros_packages/isar_turtlebot/maps/' with the name 'new_world.*'. To add a default configuration for the initial pose and position of the robot in the simulation, add '/config/new_world.cfg' with the desired parameters.

Running custom model

To run the simulation with the custom model, set WORLD_NAME=new_world in docker_scripts/entrypoint.sh.

Simulation without docker

Simulator installation

The simulator requires a computer or a virtual machine running Ubuntu. The turtlebot simulator works with different ROS/gazebo distributions, but this installation guide is based on ROS noetic. To install and run the turtlebot simulator, you will need to install ROS noetic on your computer. The desktop version should be sufficient.

Then, install dependent ROS packages, and the simulation package.

You also need to install gazebo with a version corresponding to your ROS distribution:

sudo apt-get install gazebo11

The installation can be verified by running:

gazebo

Install the required ROS packages for gazebo:

sudo apt-get install ros-noetic-gazebo-ros-pkgs ros-noetic-gazebo-ros-control

Running the simulation

First, you will need to set the following environment variables:

export ISAR_TURTLEBOT_PATH=<path/to/isar-turtlebot>
export TURTLEBOT3_MODEL=waffle

Build with catkin_make:

cd ~/catkin_ws && catkin_make

You will need a map (map.pgm and map.yaml). The map can be generated, or the example map can be used.

The gazebo together with the navigation stack and rosbridge is then launched by:

roslaunch launch/simulation.launch

The robot can be located with 2DPose Estimate and naviagted with 2D Navigation Goal in rviz.

Simulations with ISAR

Download and install ISAR on your computer. Follow the robot integration guide installing isar-turtlebot. Remember to set the robot directory environment variable:

export ISAR_ROBOT_PACKAGE=isar_turtlebot

For the ISAR API to communicate with the simulator, you will need rosbridge:

sudo apt-get install ros-noetic-rosbridge-suite

You can now run the simulation and launch rosbridge:

roslaunch rosbridge_server rosbridge_websocket.launch

To run ISAR:

python main.py

Missions can be posted to the robot through ISAR.

If the example map are used, you can try the example mission number 2.

Teleoperation

The turtlebot base can be controlled manually by publishing to the ros topic /cmd_vel. Using a keyboard, the ros package teleop_twist_keyboard translates keyboard inputs to ros messages. Similarly, the ros package teleop_twist_joy handles joystick messages.

Keyboard

With the simulation running, open a new terminal. If running in docker, access the docker container with:

docker exec -it isar_turtlebot bash

Install the teleoperation package (if running in docker this package must be installed every time):

sudo apt-get install ros-noetic-teleop-twist-keyboard

With the teleoperation package installed, enable it with:

rosrun teleop_twist_keyboard teleop_twist_keyboard.py

Joystick

Teleoperating with a joystick requires an additional package, joy, for reading the joystick input and publishing it to a topic. This package should be compatible with any joystick supported by Linux. After you connected the joystick it should be found as an input decvice /dev/input/jsX, where X is the unique id-number of the joystick (default js0).

To verify if the joystick is setup correctly, find the unique id-number with

ls /dev/input/

and run

jstest /dev/input/jsX.

To enable teleoperation with a joystick while running in docker the joystick must be added as a device in docker-compose.yaml. This is done by including the docker-compose-gamepad.yaml file as decribed in the docker section. Additionally the environment variable TELEOP_CONTROLLER must be specified(currently "xbox" is the only supported controller):

TELEOP_CONTROLLER="xbox" docker-compose -f docker-compose.yml -f docker-compose-gamepad.yml up --build

To enable teleoperation while running locally, first install the two packages:

sudo apt-get install ros-noetic-joy  ros-noetic-teleop-twist-joy

Enable teleoperation by opening two new terminals and start up joy_node and teleop_node:

rosrun joy joy_node

rosrun teleop_twist_joy teleop_node

To control the turtlebot, hold in enable_button and use the joystick. See documention for teleop_twist_joy for more information.

Simulation with manipulator

The turtlebot can be equipped with a manipulator which also can be included in the simulations. Additional software must be installed for the manipulator simulation.

cd ~/catkin_ws/src/ &&
git clone https://github.com/ROBOTIS-GIT/turtlebot3_manipulation.git &&
git clone https://github.com/ROBOTIS-GIT/turtlebot3_manipulation_simulations.git &&
git clone https://github.com/ROBOTIS-GIT/open_manipulator_dependencies.git &&
cd ~/catkin_ws && catkin_make

The manipulator can be controlled using rviz or a simpler GUI which is enabled by setting the roslaunch argument MANIPULATOR_GUI to "rviz" or "simple" respectively. With the latter as default value. There is no constraints for running both controllers simultaneously, but such functionality is not implemented. Running simulation with manipulator can be done by the roslaunch command with the prerequisite of having isar-turtlebot installed as a ros package.

roslaunch isar-turtlebot turtlebot_manipulator.launch open_manipulator_gui:=true

Alternatively the simulation with manipulator can also run in docker by including the parameter ENABLE_MANIPULATOR and the controller GUI set according to the description above:

sudo ENABLE_MANIPULATOR=true MANIPULATOR_GUI="rviz" docker-compose up --build

The simulation can also run in docker as described in the section for docker.

Manipulator teleoperation

Teleoperating the manipulator during simulation is enabled with the roslaunch argument TELEOP-CONTROLLER specifying the controller (currently 'xbox' is the only supported controller).

roslaunch isar_turtlebot turtlebot_manipulator.launch teleop_controller:="xbox"

To run simulation in docker with teleoperation of manipulator:

sudo ENABLE_MANIPULATOR=true MANIPULATOR_GUI="rviz" TELEOP_CONTROLLER="xbox" docker-compose -f docker-compose.yml -f docker-compose-setting.yml  up --build

Simultaneous teleoperation

A custom node is included the isar_turtlebot ROS package that allows for teleoperation of the turtlebot base and manipulator simultaneously. The manipulator responds directly to input and can be used without the motion planner in RViz.

The arm is controlled by moving a reference point for the end effector based on control inputs, and then using the moveit package for performing inverse kinematics at every change. The joint states are sent directly to the control system on the turtlebot.

Run the node with the following command:

roslaunch isar_turtlebot turtlebot_manipulator_teleoperation.launch teleop_controller:="xbox"

Add the argument below to activate the node without waiting for a mission step from ISAR:

roslaunch isar_turtlebot turtlebot_manipulator_teleoperation.launch teleop_controller:="xbox" teleop_activate_from_start:="True"

The control mappings are as follows:

Keyboard:

• WASD: Wheel velocities
• IJKL: Move reference of arm end effector in the horizontal plane
• UO: Move reference of arm end effector along the vertical axis
• ,.: Close and open the gripper
• SHIFT: Hold to double speed
• h: Reset to home position
• p: Pick up object
• v: Place down object
• t: Throw held object
• ENTER: Deactivate teleoperation

Xbox one controller:

• Left-stick: Wheel velocities
• Right-stick Move reference of arm end effector in the horizontal plane
• Triggers: Move reference of arm end effector along the vertical axis
• Bumpers: Close and open the gripper
• A: Hold to double speed
• Y: Reset to home position
• B: Pick up object
• X: Place down object
• View: Throw held object
• Menu: Deactivate teleoperation

Development

For local development, please fork the repository. Then, clone and install in the repository root folder:

git clone https://github.com/equinor/isar-turtlebot
cd isar-turtlebot
pip install -e ".[dev]"

Contributing

We welcome all kinds of contributions, including code, bug reports, issues, feature requests, and documentation. The preferred way of submitting a contribution is to either make an issue on github or by forking the project on github and making a pull requests.

We write our commit messages according to this guide.