Platform for data acquisition and training with WidowX in docker container.

'docker' folder - contains Dockerfile and docker-compose.yml file.
'widowx_envs' folder - docker container-mounted python package.

The following environment variables are used and can be adjusted to mount given directory in the docker container:

• CODE - mounted as '~/code' folder (defaults to 'code' folder in the repository root)
• DATA - mounted as '~/trainingdata' folder (defaults to 'trainingdata' folder in the repository root)
• EXP - mounted as '~/experiments' folder (defaults to 'experiments' folder in the repository root)

In order to make sure the default values are used and required directories are created with user ownership keep using scripts/docker-compose proxy instead of plain docker-compose when running the container.

In order to set up the host machine to run robonetv2 run:

bash scripts/host_install.sh

To avoid specifying the arm type in docker container after each container recreation, set up ROBONETV2_ARM environment variable.
For example:

echo "export ROBONETV2_ARM=wx250s" >> ~/.bashrc && source ~/.bashrc

Then run the container.

scripts/docker-compose \
   -f ./docker/docker-compose.yml \
   up \
   -d \
   --force-recreate

In order to use the provided Dockerfile.base and rebuild the image with your changes use 'docker/docker-compose.build.yml' file:

scripts/docker-compose \
   -f ./docker/docker-compose.build.yml \
   up \
   -d \
   --force-recreate \
   --build

docker exec -it robonetv2_${USER} /bin/bash

USER variable is used to avoid the clashes between the containers of different users.

scripts/docker-compose \
-f ./docker/docker-compose.yml \
up \
-d

scripts/docker-compose \
   -f ./docker/docker-compose.yml \
   down \
   --rmi all \
   --volumes

~/scripts/run.sh

For the available arguments of run.sh call ~/scripts/run.sh -h

To verify the robot functionality call the exemplary training environment:

python ~/widowx_envs/widowx_envs/widowx/widowx_env.py

After finishing the interaction before the shutdown you can call go_sleep from bash to move the robot to the sleep position.

cd ~/widowx_envs
python widowx_envs/run_data_collection.py experiments/vr_record_applied_actions_robonetv2/conf.py

You change data collection parameteres in experiments/vr_record_applied_actions_robonetv2/conf.py

1. Put the repository into the folder that is specified by the $CODE environment variable (local outside docker). Inside the docker/Dockerfile.base find the line:

put RL repo on python path, add more repos if necessary. Note that you need to set the correct CODE environment variable on the host machine RUN echo 'export PYTHONPATH=${PYTHONPATH}:/home/robonetv1/code/railrl-private' >> ~/.bashrc

Add a similar line for you new codebase.

2. Add python requirements to requirements.txt
3. rebuild the docker container

Each WidowX environment is reading 'widowx_envs/global_config.yml' created at the first run of 'scripts/run.sh'. That's a file which can be used as set of parameters which will be used by all WidowX envs (including teleoperation). For example, this file can be used to rotate the workspace in all WidowX environments on the specific machine by adding this line:

workspace_rotation_angle_z: 1.57

This configuration change will be also used by the go_sleep function. In order to avoid the changes in the configuration file from showing in git, please call:

git update-index --assume-unchanged widowx_envs/global_config.yml

To revert it, call:

git update-index --no-assume-unchanged widowx_envs/global_config.yml

git update-index does not propagate with git, and each user will have to run it independently.

The realsense cameras require different drivers than RGB cameras. If you are using realsenses, change the camera_string in scripts/run.sh to realsense:=true.

You will also need to update the device IDs in /widowx_envs/widowx_envs/widowx/launch/realsense.launch to match your cameras.

If you run into following errors:

Traceback (most recent call last):
  File "urllib3/connectionpool.py", line 677, in urlopen
  File "urllib3/connectionpool.py", line 392, in _make_request
  File "http/client.py", line 1277, in request
  File "http/client.py", line 1323, in _send_request
  File "http/client.py", line 1272, in endheaders
  File "http/client.py", line 1032, in _send_output
  File "http/client.py", line 972, in send
  File "docker/transport/unixconn.py", line 43, in connect
PermissionError: [Errno 13] Permission denied

that can be fixed by running the following commands and subsequently restarting the PC (the log out and log back in is sometimes not sufficient):

sudo groupadd docker
sudo usermod -aG docker $USER