Sawyer Forward Kinematics ML Model

This will train a model to predict forward kinematics from data acquired from Sawyer robotic arm.

Instructions to setup

• ssh to GPU machine: ssh -XY [email protected]
• On the heracleiadl machine, clone the repository and cd to it:
• git clone https://github.com/cloudy/sawyer_fk_model.git && cd sawyer_fk_model
• If you already have the repository, use git pull to update locally
• Create a virtual environment: mkvirtualenv --python=/usr/bin/python3 YOURENVNAME
• Install packages with pip: pip install -r requirements.txt
• Now train: python ForwardModelTrain.py
• For generalized 7D0F: python GeneralizedFKTrain.py

Adding data

If attempting to use data that is not presently on heracleiadl.

You can scp your data files over from your current machine.

For example, you have file1.txt stored in /home/username/dir,

and you want to move it to /home/username/dir2/dir4/ on heracleiadl.

The scp command to do this is scp /home/username/dir/file1.txt [email protected]:/home/username/dir2/dir4

Using screen

On the heracleiadl machine, train in screen so you can check progress at will and let it continue running once you end your ssh connection.

Instructions:

• ssh -XY [email protected]
• set up new screen session: screen -S YOURSCREENNAME
• This will launch the screen session, in it you can start training.
• Example: workon nntest && python GeneralizedFKTrain.py
• To exit screen without killing it, press this keyboard combo: Ctrl+a+d
• To reconnect to a screen session: screen -x
• If there are multiple sessions available, screen -x SESSIONIDNUMBER
• To exit a screen session (kill it), enter screen session and enter: exit.