jietan / puppersim Goto Github PK

After following the procedure to connect to the robot, I am not able to deploy it. I am attaching the terminal log for your reference.

% ./deploy_to_robot.sh python3 puppersim/puppersim/pupper_ars_run_policy.py --expert_policy_file=puppersim/data/lin_policy_plus_latest.npz --json_file=puppersim/data/params.json --run_on_robot PI addrees is 192.168.1.100 ssh: connect to host 192.168.1.100 port 22: Operation timed out

When calling pip install -e . and getting all the install requirements for Puppersim we run into some issues, code shown below. This stems from the new gym as of now 26.2 treating the registry module differently: openai/gym#3097.

Hotfix:

pip install gym==0.21.0  # roll back to previous version
pip install protobuf==3.20.0
!pip install importlib-metadata==4.13.0

Even after you apply the hotfix, two packages were missing:
packaging and six.

Morning,

I was working on pupper robot and I was wondering if is available the URDF/XACRO with the real meshes from the CAD model (the updated one called "Pupper 2.step").

Thank you very much,

Stefano

It's hard to compare training runs when the clipping value for the maximum speed changes. Maybe we can normalize the loss against the maximum value.

How are the actions mapping with Puppet motors? As there are 12 joints, I am not able to map the actions and observations. The ac_dim is 16 and ob_dim is 18.

What kind of experiment is best to compare the real life PD gains to the sim pd gains?

@jakegrigsby

• link a video to running policy in simulation
• add options to env to choose torque control
• add unit test that makes sure ddpg runs and doesn't crash

@jadenvc