real-stanford / universal_manipulation_interface

Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robots
https://umi-gripper.github.io/
MIT License
691 stars 135 forks source link

Question: Hardware Execution Latencies #31

Open bjin-bdai opened 7 months ago

bjin-bdai commented 7 months ago

From the associated paper, it seems that hardware latencies can play a big part in the overall performance of the policy. I noticed that in eval_real.py there is a action_exec_latency variable that is used for filtering actions inferenced from the policy. At the same time, there is a separate latency value robot_action_latency, that is used in bimanul_umi_env.py. What is the difference between these two latency values? Should they be the same value?

Bailey-24 commented 7 months ago

hope it can develop UMI2 for easy use and assemble. and it can use "Scaling Up and Distilling Down Language-Guided Robot Skill Acquisition " for language control