Видео

good job

This is excellent. The passive data presentation was very helpful.

Thanks.

Seriously? Are you guys out of your minds? That's a robot without any restriction on it's motion with a sharp EOA attachment (your fork) and you're not even using safety glasses to protect your eyes?

❗ 𝓹𝓻𝓸𝓶𝓸𝓼𝓶

I'm confused by how you get h to transition between s to s' before training the aligning model. How exactly does the data collection of getting the human input work? so to go form s to s' using h, is h being passed through the untrained transformation function in the paper, or is h being initially handled by some other teleoperation control scheme which is then mapped to the transformation map?

First!

the main issue taht limits the suit is constant misalignemnet of the joints and the lack of a few major axis of movement (leg rotation and abduction for example). this creates a frame that forces the wearer to move unnaturaly and output that can not match human movement by not being able to move like a human. a more anatomicaly correct design (better ankle and hip joint alignment, double joint knee, leg abductor behiind the back and rotator beside it, lengt adjustable upper and lower leg) would greatly improve movement fluidity and accuracy, allowing for more useful software output. a system of bungee cords to simulate the elastic properties of muscle wihout having it rely on beyond peak motor function might help too.

Hi. Where did you publish your codes related to the study?

So in the paper it says that only 2DOF joystick control is used, is that per task? It seems that the user still presses with their right hand sometimes. Is that for switching tasks or different tasks have different controls?

great talk, zhangjie !