• Introduction
• Prerequisite
• Installation
• Container
• Run
• Troubleshooting

Introduction

This is a self-contained package aimed at running the code developed for our paper ``A Shared Control Approach to Safely Limiting Patient Motion Based on Tendon Strain During Robotic-Assisted Shoulder Rehabilitation" using Docker. This should save you from having to install manually all the dependencies, and allow you to run the code in simulation (with Gazebo) as well as on the KUKA robot.

Prerequisite

You should have a working Ubuntu distribution ready on your computer (e.g. Ubuntu 22.04 LTS). Please follow the instruction on Docker and install Docker.

Installation

Go to your home folder and install this repo by

git clone git@github.com:itbellix/biomechanical_safe_deflection_docker.git
cd biomechanical_safe_deflection_docker

Follow the instructions on cloning the repos to use the controllers. You don't need to build any of the packages, the dockerfile will build them as it builds the docker image. However, please make sure to switch to the correct branch on each repo.

For kuka_fri (used for the real robot). If you don't have access to it, you can skip the next instruction block, and still work in simulation.

mkdir kuka_dependencies
cd kuka_dependencies
git clone --recurse-submodules git@gitlab.tudelft.nl:kuka-iiwa-7-cor-lab/kuka_fri.git
cd kuka_fri
git checkout legacy
cd ..

For other repos

git clone --recurse-submodules git@github.com:costashatz/SpaceVecAlg.git
git clone --recurse-submodules git@github.com:costashatz/RBDyn.git
git clone --recurse-submodules --recursive git@github.com:costashatz/mc_rbdyn_urdf.git
git clone --recurse-submodules git@github.com:mosra/corrade.git
cd corrade
git checkout 0d149ee9f26a6e35c30b1b44f281b272397842f5
cd ..
git clone --recurse-submodules git@github.com:epfl-lasa/robot_controllers.git
cd ..

Now you can build you docker image by

docker build -t biomech_safe_defl .

this can take a while, grab a coffee, make some tea, and after image is finished building, you can check your list of available docker image by

docker image list

You should see information about the image biomech_safe_defl which we just built!

In the docker file, we created a user kuka_cor with USER_UID and USER_GID both at 1000, same as your host. This mean that you can share volumes and files between docker image and host.

Container

Here, you create the actual docker container that will allow to run the code. You should put packages you wish to run on kuka into the catkin workspace's src. This always includes the TU Delft implementation of iiwa_ros. Here, we will also install the biomechanical_safe_deflection package, and check out to a tested version of it. Clone and set up the packages by

cd catkin_ws
mkdir src && cd src 
git clone --recurse-submodules git@gitlab.tudelft.nl:kuka-iiwa-7-cor-lab/iiwa_ros.git
git clone --recurse-submodules git@github.com:itbellix/biomechanical_safe_deflection.git
cd biomechanical_safe_deflection
git checkout d095012ca9710291d97e7785af14234eeaeeaa18
cd ../../..

If you are working only in simulation (i.e., you don't have access to the kuka_fri package), run the following:

cd catkin_ws/src/iiwa_ros/src/iiwa_driver
touch CATKIN_IGNORE
cd ../../..

Run the docker image and load directory iiwa_ros as an volume

docker run -it --user kuka_cor --name my_container --network=host --ipc=host -v $PWD/catkin_ws:/catkin_ws -v /temp/.X11-unix:/temp/.X11-unix:rw --env=DISPLAY biomech_safe_defl

This will create a container named my_container with the image kuka_ros, and at any point, you can exit the container with ctrl+ d.

Note: the first time you run this you should see this error bash: /catkin_ws/devel/setup.bash: No such file or directory. This is because we have not run catkin build yet, so don't worry!

You can start it again with

docker start -i my_container

or you can also open up a different terminal in the same container by

docker exec -it my_container bash

Firstly, you should build your working package by

cd catkin_ws
catkin build
cd ..
source catkin_ws/devel/setup.bash

Any changes made by catkin_make will also show up in the host directory so you don't need to rebuild the package every time you rerun your image. And sourcing of your package is taken care of in bashrc.

Run

Open 2 terminals in the container, and :

1. the robot controller (and Gazebo if simulation argument is set to true)

roslaunch biomechanical_safe_deflection controller.launch simulation:=true

2. the shared controllers

roslaunch biomechanical_safe_deflection start_all.launch simulation:=true

You should be see Gazeb firing up with a KUKA LBR iiwa 7 model, and a dynamic_reconfigure window that allows to select the task to execute.

Note 1: we do not ship the HSL libraries necessary to run the MA27 solver with IPOPT. As such, the default solver is MUMPS, which achieves similar performances but is a bit slower. You can always build the HSL libraries from source yourself, and use a linear solver of your choice.

Note 2: likely, the scripts running in the docker container will not have access to the audio drivers of your device. If this is the case, you will see a warning being printed during the high-authority interventions instead of hearing a sound as described in our paper.

Note 3: our controllers are designed for actual human-robot interaction, so the simulation setup is quite minimal and allows only to have a rough idea of the overall functioning of our system... For example, when estimating online the position of the center of the human shoulder, the simulation will gradually drift and accumulate errors, so it appears as if the robot is moving away. If you want to have a better sense of the desired results, check out our video or try connecting your robot!

Troubleshooting

Nvidia Issues [Docker]

If you are having problem using your Nvidia GPU in docker GUIs, please take a look at Nvidia Container toolkit and follow their instruction. Generally, adding args into the docker run command helps, e.g. --runtime=nvidia and --gpus all. For example, the run command will be like

    docker run -it --runtime=nvidia --gpus all --user kuka_cor --name my_container --network=host --ipc=host -v $PWD/catkin_ws:/catkin_ws -v /temp/.X11-unix:/temp/.X11-unix:rw --env=DISPLAY kuka_ros

Missing GPU [Docker]

If your system does not have a GPU, you will see an error like the following when trying to run some applications (e.g. Gazebo):

libGL error: MESA-LOADER: failed to retrieve device information
Segmentation fault (core dumped)

You should make sure that OpenGL uses software rendering instead, by running:

export LIBGL_ALWAYS_SOFTWARE=1

You need to run this in every terminal, so add this line to the ~/.bashrc in your container, to set it on every terminal session.

ROS Service Issues [Docker]

If ROS service is having trouble to run, a possible cause this that RBDyn can't be found. Run this line in terminal and try again:

    export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib:/kuka_dependencies/RBDyn/build/src