Multiple Repositories, many moving parts not clear where to start with Imitation learning

[ZacharyDK]

Saw the videos of the mobile aloha project, and wanted to start dissecting it for my own learning. However, it seems there are multiple layers in which aloha is built from.

Since yall are the foremost experts on Imitation learning, I'd appreciate if you could provide some direction about getting started. Goal: Get a 3D virtual model of a simple robotic arm to work with imitation learning.

There are the important base components so far, not sure if I'm missing anything.

Mobile ALOHA: comes from ALOHA
ALOHA: Comes from ACT.

ALOHA: repo provides a link to the act repo
Mobile ALOHA: relies on act plus plus repo. Website

ACT Plus Plus - linked to from website under the ML link.

Two Papers. One related to ACT, the other related to Mobile ALOHA.

"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware" paper here

"Mobile ALOHA: Learning Bimanual Mobile Manipulation with Low-Cost Whole-Body Teleoperation" paper here

Also the interbox module seems to be used a lot