Stopped after `sudo make autostart`

[Kashu7100]

Thank you for sharing the code for this awesome work.

I followed the setup in the README to set up your code and I would like to have some advice on the Running the Controller section. After running the second command (cd ~/go1_gym/go1_gym_deploy/docker; sudo make autostart), the code seems to stop after a while (please refer to the attached image).

I pressed [R2] several times at this moment but it did not do anything.

Do you have any suggestions about this situation?

Sincerely,

[gmargo11]

Hi @Kashu7100 ,

Great to see you're trying out the code on your Go1!

To diagnose the issue, rather than using the autostart script, let's enter the container and start the controller manually. Can you let me know the output of the following commands? First instead of the autostart command:

cd ~/go1_gym/go1_gym_deploy/docker; sudo make run

And then, if that command successfully enters the docker container:

python3 setup.py install
cd go1_gym_deploy/scripts
python3 deploy_policy.py

Gabe

[Kashu7100]

After python3 deploy_policy.py command, the process seems to stop at pkl_cfg = pkl.load(file)

[Kashu7100]

After reading your paper, I noticed that my Go1's jetson board is normal nano (not NX). I think maybe the lack of computation power of my Go1 is causing this issue.

Do you think running without docker can help? Also, I want to check whether I can run the policy from my PC (not the onboard nano). I am wondering if there is a way to communicate with lcm_position from outside of Go1 via ethernet cable.

[gmargo11]

Odd! But the parameters.pkl file is present at /home/isaac/go1_gym/runs/gait-conditioned-agility/pretrain-v0/train/025417.456545/parameters.pkl, right?

You could try ensuring this file is fully readable by all users by navigating to runs/gait-conditioned-agility/pretrain-v0/train/025417.456545/ and running chmod +777 parameters.pkl

Re: Other questions

• The controller should run fine on a normal Nano, NX is not necessary. Also, this line is just loading a small file using the pickle library which shouldn't be computation-intensive.
• You can run the whole controller outside docker, but you would need to install the dependencies on your Nano manually. Docker is easier for sharing code across systems. A simpler debugging step might be to first ensure that this pkl.load command can run successfully outside docker.
• I haven't tried networking an external PC with the Go1. But it's definitely possible. The deployment script communicates with lcm_position by sending and receiving LCM messages. So you just need to ensure that the LCM messages published by your PC are received on the Go1. You might find some information here: https://lcm-proj.github.io/

[Kashu7100]

Thank you for your advice. Actually, chmod +777 parameters.pkl works. After the chmod and starting the controller manually, the controller successfully works.

It seems like there are some stability issues with running the code, but I will look deeper! Also, it would be helpful if you can publish the source code for lcm_position file.

Thank you a lot.

[gmargo11]

Great to hear that resolved the issue. I've just made the file permissions correct in the repo. I also added the source code for the lcm_position file.

Closing this issue since I think we've eliminated the original error. If you open a new issue with any stability problems you're referencing, happy to take a look.

Also, would love to see video if you take the system out for a successful walk :-) Thanks for being an early tester of the deployment code

[Kashu7100]

Thank you for adding the source code for lcm_position.

It seems the following header files are required to compile:

#include "state_estimator_lcmt.hpp"
#include "leg_control_data_lcmt.hpp"
#include "pd_tau_targets_lcmt.hpp"
#include "rc_command_lcmt.hpp"

Some of them seemed like from the Cheetah Software but I could not figure it out.

Also, I will definitely post the video!

[gmargo11]

Here's a fork of the unitree_legged_sdk with all of the files needed to build lcm_position: https://github.com/Improbable-AI/unitree_legged_sdk