This README is the definitive source for downloading, installing, and running the Personal Robotics Lab's robot-assisted feeding software. This code has been run and tested on machines running Ubuntu 22.04 and ROS2 Humble.
-
Install ROS2 Humble (binary packages are recommended over building from source), configure your environment, and create a workspace.
- NOTE: In the remainder of the instructions, replace
~\colcon_wswith the path to your workspace.
- NOTE: In the remainder of the instructions, replace
-
Configure
pr-rosinstallsin order to download all necessary repositories.cd ~ git clone https://github.com/personalrobotics/pr-rosinstalls.git -
Pull all the repositories, with the correct branch. Replace
<https/ssh>with one of the options, depending on your authentication method.sudo apt install python3-wstool # if not already installed cd ~/colcon_ws/src wstool init wstool merge ~/pr-rosinstalls/ada-feeding.<https/ssh>.rosinstall wstool up -
Configure
rosdep:sudo apt install python3-rosdep # if not already installed sudo rosdep init # if this is the first time using rosdep -
(Only for users with sudo access) Install
rosdepdependencies:rosdep update cd ~/colcon_ws rosdep install --from-paths src -y --ignore-src --as-root=pip:false -
Install and fix pip/non-
rosdepdependencies:cd ~/colcon_ws/src/ada_feeding python3 -m pip install -r requirements.txt-
Upgrade
transforms3d, since the release on Ubuntu packages is outdated:python3 -m pip install transforms3d -U -
Remove the duplicate matplotlib pip installation caused by installing scikit-spatial with pip (some accounts have required sudo access for this command, other have not. If you do not have sudo access and encounter this, contact a lab member who does):
python3 -m pip uninstall matplotlib -
pyrealsense2is not released for ARM systems, so ARM users will have to build from source. You may have to add the-DPYTHON_EXECUTABLE=/usr/bin/python3flag to thecmakecommand. When runningsudo make install, pay close attention to which pathpyrealsense2is installed to and add that path to thePYTHONPATH-- it should be/use/local/libbut may be/usr/local/OFF.
-
-
Install the JACO SDK (real robot only). All SDKs are listed here; PRL currently uses the Gen2 SDK v1.5.1. Note that although the latest version of that SDK is for Ubuntu 16.04, it still works on Ubuntu 22.04 (only for x86 systems, not ARM system).
-
(Optional): We recommend using CycloneDDS as your ROS2 midleware, with a custom configuration file that enables multicast on loopback and prioritizes loopback. Install it with
sudo apt install ros-humble-rmw-cyclonedds-cpp. Then, add the following lines to your~/bashrc:export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp; export CYCLONEDDS_URI=~/colcon_ws/src/ada_feeding/cyclonedds.xml.- Note that you may have to change the name of the fallback interface that CycloneDDS uses if
lodoes not work for that transmission. To do so, runifconfigand either use the name of your ethernet network or WiFi network, depending on how your computer is connected to the internet. - Note that you have to rebuild your workspace from scratch after doing this.
- Note that you may have to change the name of the fallback interface that CycloneDDS uses if
-
Build your workspace:
cd ~/colcon_ws colcon build --symlink-install # if sim-only, add '--packages-skip ada_hardware'
Our system includes two computers, lovelace (Lenovo Legion 5i) and nano (Nvidia Jetson Nano), where the latter runs camera code and the former runs everything else. This repository includes some system-specific bash scripts to configure those computers. Specifically, move start_nano.sh, start_nano.py, run_camera.sh, and run_nano_bridge.sh to the home directory of nano. Copy nano_bridge into the ROS2 workspace on nano (e.g., ~/Workspace/camera_ros2), and rebuild it. On lovelace, add a cronjob to run configure_lovelace.sh on reboot (e.g., sudo crontab -e, then add the line @reboot <path/to/ada_feeding>/configure_lovelace.sh).
Note that these scripts will likely need to be changed for your setup's computers; however, perhaps the commands in these scripts can provide you inspiration.
If you change the e-stop button, the e-stop button's adapter(s), and/or the device this code is being run on, you may have to change several device-specific parameters for the e-stop to work. Read the long comment above amixer_configuration_name in ada_feeding/config/ada_watchdog.yaml for detailed instructions on changing these parameters.
-
Install the Node Version Manager (nvm): https://github.com/nvm-sh/nvm?tab=readme-ov-file#install--update-script
-
Install and use NodeJS 21 (Note: if you have just installed nvm using the previous command, you will need to source your .bashrc or open a new terminal to run these commands):
nvm install 21 nvm use 21 -
(Only for users with sudo access; this should already be configured on PRL computers) Make Node available to all users, including root:
sudo ln -s "$NVM_DIR/versions/node/$(nvm version)/bin/node" "/usr/local/bin/node" sudo ln -s "$NVM_DIR/versions/node/$(nvm version)/bin/npm" "/usr/local/bin/npm" sudo ln -s "$NVM_DIR/versions/node/$(nvm version)/bin/npx" "/usr/local/bin/npx" -
(Only for users with sudo access; this should already be configured on PRL computers) Install
serveandpm2globally. Root access is necessary forserveso it can access port 80.sudo npm install -g serve npm install -g pm2@latest -
Install the web app dependencies. (Note: there will be some vulnerabilities in dependencies. This is okay, since access to the web app is shielded behind our secure router.)
cd ~/colcon_ws/src/feeding_web_interface/feedingwebapp npm install --legacy-peer-deps npx playwright install -
(Optional; this should already be configured on PRL computers) To access the web app on a device other than the one hosting it, enable the desired port for HTTP access: https://www.digitalocean.com/community/tutorials/ufw-essentials-common-firewall-rules-and-commands#allow-all-incoming-http-port-80
We use the convenience script start.py to launch the software. This script has several command-line arguments, which can be seen by passing the -h flag when running the script.
This option starts the web app and the real robot code, and can be used to test the entire system. This will by default start the web app on port 80, and requires sudo access. The robot's HDMI connection must be plugged into your computer, and your computer should be connected to the robot's router with an ethernet cable.
NOTE: If not running on the production machine i.e., lovelace, it's recommended that you append the command line flag --dev to the start script. This will launch RVIZ, will not require the e-stop button to be plugged in, and will not require sudo access to launch the web app.
cd ~/colcon_ws
python3 src/ada_feeding/start.py
In a browser, access 127.0.0.1 (if on the same device serving the web app), or the IP address of the device hosting the web app (if on a different device connected to the same network, e.g. a cell phone connected to the LOVELACE_5g network). You should now be able to run the system! Note that upon startup, the watchdog is in a failing state until the e-stop is clicked exactly once, allowing the system to verify that it is connected and working.
To close, run python3 src/ada_feeding/start.py -c
This option starts the web app, runs dummy nodes for perception, runs the real robot motion code, but runs a mock robot in MoveIt. This is useful for testing robot motion code in simulation before moving onto the real robot. This will start the web app on port 3000 and does not require sudo access.
cd ~/colcon_ws
python3 src/ada_feeding/start.py --sim mock
In a browser, access 127.0.0.1:3000 (if on the same device serving the web app), or the IP address of the device hosting the web app at port 3000 (if on a different device connected to the same network). You should now be able to run the system!
To close, run python3 src/ada_feeding/start.py --sim mock -c
This option starts the web app, and runs dummy nodes for all perception and robot motion code. This is useful to test the web app in isolation. This will start the web app on port 3000 and does not require sudo access.
cd ~/colcon_ws
python3 src/ada_feeding/start.py --sim dummy
In a browser, access 127.0.0.1:3000 (if on the same device serving the web app), or the IP address of the device hosting the web app at port 3000 (if on a different device connected to the same network). You should now be able to run the system!
To close, run python3 src/ada_feeding/start.py --sim dummy -c
- Install
pre-commit:python3 -m pip install pre-commitcd ~/colcon_ws/src/ada_feedingpre-commit install
- Before opening a pull request:
cd ~/colcon_ws/src/ada_feedingpre-commit run --all-filespylint --recursive=y --rcfile=.pylintrc .
- Something is not working, what do I do?: All our code runs in
screensessions, so the first step is to attach to individual screen sessions to identify where the issue is. Runscreen -lsto list all screens, runscreen -r <name>to attach to a screen session, andCtrl-A dto detach from the screen session. An introduction toscreencan be found here. - The watchdog is not recognizing my initial e-stop click: Verify the e-stop is plugged in, and that any other processes accessing the microphone (e.g., Sound Settings) are closed. Run
alsamixerto see if your user account has access to it. If you do not see sound levels in the terminal, trysudo alsamixer. If that works, give your user account permission to access sound:sudo setfacl -m u:$USER:rw /dev/snd/* - The watchdog is failing due to the F/T sensor: First, check whether the force-torque sensor is publishing:
ros2 topic echo /wireless_ft/ftSensor3. If not, the issue is in theftscreen (one potential issue is that the aliasft-sensordoes not point to the right IP address for the force-torque sensor, in which case you should pass the IP address in using thehostparameter). If so, check the timestamp of the published F/T messages compared to the current time. If it is off, the problem is that NTP got disabled on the force-torque sensor. You have to useminicomto re-enable NTP (see here for PRL-specific instructions). - I get the error
cannot use destroyable because destruction was requested: Upgrade to the latest version ofrclpy. - I get the error
bad option: --env-file=.envwhen launching the WebRTC Signalling Server: You are using an old version of Node; upgrade it to 21.