วีดีโอ

Imagine it's optimized to the point where it can essentially toss/flick this thing, spinning it, and then catching in all under half a second. Then imagine these types of robotics not taking every single manual labor job in the next 10 years...

WoW great!

Thanks for your sharing! But as a freshman, I didn't know whether the process data is from the physical sensor in the real world (i.e. cyber gloves) or just simulated by the computer? If the data is from the real sensors, how to realize the transfer the signal data from real sensor to the Mujoco Haptix program? Thank you!

Good morning

fantastic, will you release the code and when?

sir, Do you have any paper reference for this ? it would be helpful for me.

This is so exciting

This job is so great !👍

Please add the XML files soon :D Excellent work!

cool stuff vikas

Bahut ache

Veri nice

Hanzhen harmonic drive gear , over 30 years experience , robot arm gear

Hanzhen harmonic drive gear , over 30 years experience , robot gear

Hanzhen harmonic drive gear , over 30 years experience , robot arm gear

god

great

Very cool work

Awesome work,vikash...

Hey Vikash! I've looked through the documentation and it seems like the servos are the most expensive element for the robots. Do you know of any affordable alternatives to the XM430-W210-R?

Planning to kill some humans humm mate?

inspiring work! questions: arxiv link? github? was this all done via training in simulation and then transferring to hardware? or was some training done in a physical environment? also is there any benefit to say train in a physical environment exclusively with sensor feedback - or does simulation always end up being cheaper? or are hybrid methods the way to go? do these models or policies break down if say the motion were sped up several orders of magnitude? 10x, 100x? one assumes that the physics themselves would be slightly different so the policy unless robust to changing physics may or may not be able to handle that. - but it's cool seeing deep learning methods applied to fine precision motion/dextrous robotics, the open-ai hand being another cool example of which

where can i get the software?

...хотя это, в основном, лишь компьютерная анимация..., впечатляет...!!!

Amazing. Hmm. I wonder if the hand could be used to...never mind. 😁

Loved this! The LEGO style design is very impressive. Are the CAD/code files for this project open source? Would love to make my own so I can do some research on myoelectric control

Which simulation platform have you used?

I really can't wait. Please freeze me and wake me up when we have robots. Good job guys.

Very cool. Congratulations!

Cool video :D. Keep it up!

Im glad ur teaching it how to give a hand job early on, it will master it soon

OK..this baby cant feed itself yet. KEEP WORKING ON IT ! Skywalker needs a hand..

BRAVO guys. Очень круто.

So if I wanted it to grasp a teaspoon at different positions and orientations from the table, lift it up, and bring it into the stirring posture with a few re-grasps, I would have to: 1) apply some infrared markers to the spoon 2) handcode an appropriate loss function in Matlab 3) bring the policy near the global optimum with an initial demonstration, and 4) let it search for the local optimum on its own with ~10*5 trials, manually resetting the starting conditions after each trial?

fucking amazing