I thought so too, but real company, so appears may be real
Iām going to say take a HUGE helping of skepticism with that video.
Chinese company, unverifiable video, performance that appears to good to be trueā¦
I get it, but from everything I can find, it appears real. If you can find something that disputes that, I would be interested (and a little relieved, lol). It reminds me of the first time I saw Boston Dynamics Big Dog. At the time it looked like science fiction, of course now itās very old news.
I fail to see how the company being Chinese has any bearing on this.
The other points, sure.
This looks pretty real.
There were a couple frames where it looked like leg elements bent - they could be designed to, or those could be CGIād stunts, or my eye could be tricking me.
Theyāve had the G1 out for a while and this is a larger scale evolution of it, with bonus skate wheels. The leg stability algo is all old news from previous platforms, and throwing a onewheel under each corner is evolutionary not revolutionary from a tech perspective. Makes for really cool, perhaps revolutionary product capabilities.
(Note the video shows a series of stunts, but not the robot flowing between the different stunt modes.)
Sale price $100k for a Chinese firm is functionally around $1m for a US firm in terms of human effort purchased on the backend, to use a back-of-the-envelope conversion between the countries technical talent environments.
Unitree is a legitimate company. Iāve briefly worked with them in the past. They have some very very talented people.
Chinese tech companies with ābreakthroughā announcements and scientific advancements have a history of fabrication and exaggeration.
When objective journalists outside China can see these in action and confirm their capabilities then Iāll also congratulate them on a really neat product.
This is not a problem that is solely Chinese. But they are, frankly, known for it. Not sure why this statement of fact resulted in a thumbs down.
There is a notable difference between being prejudiced against something versus acknowledging past history and being forewarned.
I think thatās where Iām going to stop my part in this discussion regarding China.
Canāt wait to see an frc bot with this drivetrain
This could also mean it took 50 attempts to get 1 good shot of it working right. Notice how each stunt is also in a different location? How much setup is required to do that stunt in that area? Does it have to learn the space first? All of that information gets glossed over in videos like this and an end user ends up disappointed when thereās 50 little āhangupsā when you buy one and try these same things at home/work. Even then it may not ever work like you saw in a video on your different environment.
Never believe a hype video, trust testing videos with long run times or timelapses of 0 mistake runs over and over. Reliability of a skill/ability is much more important.
A video like this https://youtu.be/aLwbjCPZ958?si=Hvf7JoGwEx6-y5j_ is what a real engineer will want to see for the leg joints, wheels, etc over an extended period of use. Itās cool to see a flashy new toy, but itās not a useful machine or tool until it can be relied upon indefinitely to always succeed.
A 3 out of 5 times rule doesnāt apply here (how I judge little kids robots being successful). It needs to be 100% successful 100% of the time. Production and real world uses canāt afford a failure. In Material handling one damaged load could be thousands of dollars wasted. One injured person during a search and rescue is a million dollar lawsuit waiting to happen. We need to see proof of the reliability in action.
We have a Unitree B1 at my work.
I certainly trust that its a legitimate product, and that like all promo reels, these were not shot in 1 take.
Iām curious which of these features are actually user-facing.
The B1 does not ship with autonomy enabling this level of sophistication, though it is an open enough platform and has sufficient on-board sensing that you could program your own.
Ftfyā¦
You know about the legendary Apple iPhone demo choreography, right? The device Jobs brought onstage rebooted every fifteen seconds, so he brought it behind his back for the reboot cycle and then forward to show the next function, then hid it again before it rebootedā¦
I attended a Stockholders demonstration for a robotics company near me. We got the behind the scenes tour ahead of time and why were there was to talk about a possible partnership between our companies. So we were watching the crew trying to get the demo unit online and running behind the building 5 minutes before showtime running like chickens with their heads cutoff. They had to skip several parts of the demo/talk about it rather than show it because it wasnāt ready yet and reliable.
This is a company with a $11 million+ revenue yearly still not ready before they said they were. Obviously much smaller than Apple, but this is a common trend in Hype/Presentations Iāve seen in this industry that makes me basically assume Iām being lied to whenever I see these kinds of demos.
Thatās a bit of an exaggeration of what happenedāit was extremely prone to running out of memory (as well as all sorts of other bugs), so they spent a few months working out a super specific order (known as the golden path), that wouldnāt crash the phone. And to compensate for memory issues, they did have a few extra phones hooked up, but Iām not sure if they ever switched to any of them. They just got extremely lucky that it went off without a hitch.
The bigger point: All companies embellish their demos and do things not representative of real world scenarios, whether thatās using better hardware, taking a dozen cuts, or only showing specific angles and routines.
Unitree is indeed a real company that produces functional robots and I am absolutely blown away that they are able to get their robots down to the price point that they are.
We own a Unitree Go2 and itās extremely impressiveā¦it can even score a few points in Crescendo.
Whatās the intended use? Outreach stuff?
You could give it a harness and use it to pull the robot cart.
Wrong James Cameron sci-fi franchise from (roughly) the 1980s.
Use it to pull the robot cart or the drive team āAdam Savageā style.
But if I was attending an outreach for an FRC team, they whip out a robot dog and then I never get to use said robot dog as part of the team (i.e. we never build anything like it) Iād be extremely upset. Itās eye catching, but itās not what the team does, or builds anything like it. Same reason we didnāt keep some old Pegasus robot arm stations I inherited when I took over the program. They look cool, but if thatās the recruiting machine, it gives people the wrong impression of what you do. A full FRC robot is way cooler than an arm. Canāt say the same for the comparison to a robot dogā¦ Robot dogs are definitely cool, just be very clear that this is not what we do. Itās what you could do after FRC.
It is an amazing level of motion control and locomotion. Feels like we are at the knee of a technological inflection point.