Go2 Sim2real experience

[ammousa]

Thanks for the excellent code. I’d appreciate it if you could share your experience transferring the policy directly to the Go2 robot. Specifically:
• Did you use the original actuator network from the simulated Go1, or train a new one?
• What architecture and data volume did you use?
• Any key insights or challenges during the sim-to-real transfer?

Thanks again!

[fan-ziqi]

Sorry for the late reply

I didn’t use the actuator network. I just needed to increase the domain randomization and add delay to motors to implement sim2real well.