We are organizing discussion threads around the Unitree G1 track in SVRC Academy. What latency threshold starts to harm imitation data quality? Contex...
We are organizing discussion threads around the Unitree G1 track in SVRC Academy. What latency threshold starts to harm imitation data quality? Context: this thread is for industry labs and focuses on beginner execution. Please share practical details from your own builds. If you respond, include one concrete metric, constraint, or failure mode you observed.
Useful angle. If possible, include your hardware setup, control loop rate, and one thing you changed after the first failed attempt.
In our Unitree G1 runs, the biggest issue was consistency between sessions. We improved by standardizing pre-run checks and logging a fixed metric set.
For teams trying this next week: what is one small experiment you can run in under 2 hours to validate this advice before scaling?