The Optimus robots at Tesla’s Cybercab event were humans in disguise
The robots were just a little too slick.
The robots were just a little too slick.
You're viewing a single thread.
A robot isn't just software: the hardware for humanoid robots has proved quite difficult to build and if these robots have the hardware necessary to walk around and manipulate objects (in the real world, not in a lab where they get multiple "takes") then they are remarkable even if their every action is directed by a human.
Also I would have guessed that Hasbro owned the IP for robots named "Optimus". Maybe Tesla paid them, the way that Verizon paid Lucasfilm for the right to use the word "droid".
While it is impressive, it's impressive in a "college robotics classes were doing this six years ago" way. The control loop would be interesting with how the human gesticulation interacts with the locomotion algorithm they're using, but that assumes that the robots can walk and chew bubblegum (or in this case, walk and interact with a crowd via teleoperation) at the same time. I wouldn't be surprised to learn that the robots were entirely software controlled when walking. That seems like the case from what I'm seeing in these vids.
It has been a while since I went to college but I don't remember college robotics classes building robots capable of bipedal motion at all.
Not trying to be contrary here: have you taken any robotics courses? We do walking algorithms for our sophomore classes now, as an intro to complex multivariate simulation. Desktop sized, but motion tracking for telemetry control of end effectors is a common enough student project that we've started requiring people to modify their project proposals so they're not just implementing some online tutorial by rote. (SLAM on hexapods is another classic undergrad project, though one that doesn't have nearly as much direct applicability to the topic at hand)
I'm not trying to be contrary either - I really haven't had any exposure to this sort of thing in a long time. I'm not talking about software. Do undergrad-level classes really have access to hardware capable of bipedal motion now? I briefly looked up bipedal robots and I didn't find anything capable of walking other than the products of serious robotics companies.
Yeah, they do! The prevalence of 3D printing and cheap chinese harmonic drives has lead to a ton of open-source robotics platforms. A lot of the bipedal work we do is focused on how to construct a useful model for simulation so you don't have to watch the $20,000 humanoid robot trip down the stairs, but quite a few students have built very robust custom hardware using the uni's tools.
Do you have a link to videos of this? I'm asking sincerely.
(I'm sorry, I wrote this out then totally forgot to hit send:)
Oh totally! Here's a random sampling. None of these are quite as impressive as commercial ones, but they're all a neat intro to open-source robotics. Apologies that none of these are directly related to my classes, doxxing myself is fun and easy but also gets not fun real quick.
OSU bipedal robot, not exactly undergrad work but still really cool.
This is just a demo of a desktop robot using OpenCV, but it's really cute looking so I included it.
With how much raw physical power machines display I think people really underestimate how impressive (and thus hard to replicate) biological systems are, the human body runs on about the same amount of total energy as a mid-range laptop. If I tried to build a working humanoid robot with the same power restriction it'd be so underpowered as to be less physically capable than a young child.
