- Putting it all together for red cube picking instead of purple ball kicking
- talking about ALOHA (ACT) and LeRobot
- Steve Wozniak coffee test
- Running through what happened in November and December and plans for January

- Analogies between this and sock picking
- 4 different approaches to both: visual servoing, motion planning (classic robotics), reinforcement learning, imitation learning
- hardcoded keyfrrame for side grasp and front grasp
- first accidental almost a autonomous sock grasp

Also:

- getting up from the ground
- face tracking in camera and just waving (but no pan tilt control for this application)

Public notion page detailing my daily progress towards this and harder tasks:
bright-trollius-2c1.notion.site/Humanoid-Log...

Let's see if I can exhaust of him his possibilities. Hardware limits applications but most humanoid problems are software related which is my focus.

Therefore autonomous sock picking is the goal! 😅😜

Later I run face detection and go through all his actions. Then we SSH in, run some code to do even more actions like "hurdles", "greet", "wave", "calib", "left kick". Especially during hurdles it is clear that he is quite an impressive machine for his cost.

Also have a public notion log and plans of everything I am doing with him and related to humanoids:
bright-trollius-2c1.notion.site/Humanoid-Log...

I like this idea of building in public, keeps you accountable and also shows the progression from meagre beginnings to the State of the Art (SoTA)

In this part we go through all of his degrees of freedom comparing to human muscles + bones, cover some of his electronics and specific details and open the series.

But I think the real niche is to talk about all humanoid software, which is lacking, I will cover in the future: