Robots are about actuators, not AI

Everyone wants the universal household robot. According to Jim Anderton, for widespread adoption, they are going to have to have a price point that allows monthly financing or lease payments that are roughly similar to a car, suggesting that manufacturers will need to retail units in the neighbourhood of $ 40,000 to get widescale uptake.
If designed properly, the machines could be durable enough to carry a residual value, creating a secondary market for used equipment, to allow monthly payments that could be affordable for the majority of households.
To achieve this, the robot makers are going to have to stop thinking like NASA, and rethink things like titanium and carbon fiber. Commodity plastic resins, utility grade aluminum alloys and critically, affordable batteries will be the way forward.
* * *
ABOUT END OF THE LINE:
Manufacturing veteran James Anderton expresses his compelling and unique opinions about the state of the manufacturing sector and the engineering industry more broadly. He shares his thoughts and insights to help engineering and manufacturing professionals navigate through the challenges of world events, the blending old with new technologies, evolving processes, gaps in skilled labour, in an effort to help maximize productivity of their daily operations. Occasionally, he just likes to vent.
James is a former editor of trade publications in the automotive, metalworking and plastics industries with contributions to a wide range of print and on-line publications. He also brings prior industry experience in quality and manufacturing for a Tier One automotive supplier.
* * *
JOIN ENGINEERING.COM:
Easy, free. Join and get access to more stories, videos, white papers, research reports made for the engineering professional. Customize your profile to follow topics you care about and engage with the global engineering community. engineering.com
* * *
STEM GAMES:
Play some fun engineering-based STEM games on engineering.com:
www.engineering.com/resources...
* * *
ENGINEERING.COM TV:
Watch in-depth engineering videos made for the engineering professional and found exclusively on engineering.com TV:
Panel Expert Roundtables: www.engineering.com/viewAll?c...
For the Design Engineer: www.engineering.com/viewAll?c...
For the Manufacturing Engineer: www.engineering.com/viewAll?c...
* * *
#robots #AI #automation

[@incognitotorpedo42]

Actuators, not AI? Mm, no. It's both.

[@yapdog]

Exactly. This video is weird.

[@FLORIDIANMILLIONAIRE]

@@yapdog what to do with very advanced control system when mechanical system doesn't work well ?

[@yapdog]

@@FLORIDIANMILLIONAIRE What to do with a mechanical system without a very advanced control system? Most robotics students lean that this is the true start of their journey after they build their machines.

[@getstuk87]

Should have said "not just AI" but he clickbaits instead. Downvoting clickbbaiting is the only solution.

[@yapdog]

1:22 Self-driving cars most certainly do NOT handle much of our driving duties. What are you, reporting from 2034???

[@Dr.Kraig_Ren]

Making video future proof?

[@yapdog]

@@Dr.Kraig_Ren LOL!!!

[@ThomasTomiczek]

They do - FSD 12 current demos are insane. Not he does not say "most of the driving" but most of the driving DUTIES - i.e. where the work, they do most of the duties. Weymo has self driving car fleets in multiple cities. So, sorry, it is a correct statement.

[@yapdog]

@@ThomasTomiczek Wake me when your car picks up the kids.

[@ThomasTomiczek]

@@yapdog Keep sleeping and ignoraing the world. Smart people - so obviously not you - would look up the videos of self driving taxis in - oh, youtube. Btw., you can sleep forever - not having children, not wanting them. Also, what you just describe would not only require a self driving car, but also - well - supporting school and a lot of other stuff. Well, can not fix dump. Use youtube search.

[@pallinagialla97]

Yes, please! I'm a girl and like the vast majority of us, we have so many chores to do, and do pretty well, for our men and family. It is exhausting! Also because I do not have a bachelor's and a master's degree to clean everyone's stuff around the house 😅. As much as also chores have a dignity and so on, but please simply get them out of our way! 🙏💓

[@aware2action]

What we need is intelligent actuators, that allow for a continuous feedback. By intelligent, they should be able to do local processing, offloading compute, while reducing latency.

[@mister_r447]

Isn't unitree's new robot, G1, at 16k?

[@ThomasTomiczek]

It is also shoulder to belt height - more a toy than a seirous robot, having problems i.e. washing dishes.

[@maestrulgamer9695]

Honestly,i thought this was actually obvious! While the software evolves very quickly,and can be scaled relative easy,the hardware is much slower and harder to produce in large amounts.

[@johannesdolch]

They don't have to think like Nasa. They have to think like SpaceX. Hmm ....

[@solifugus]

For decades, I have been thinking and experimenting on this topic. AI is important but is not the problem we face today. We need low cost, highly functional, and highly durable humanoid robotics. One important principle for achieving low cost and durability is reducing the parts count. I believe PVC is the best material for this. It is inexpensive and durable and could help you do away with the need for bearings, as it is "self-lubricating" (long polymers) and has very low abrasion. As for cost and functionality, the humanoid requires many degrees of freedom and lots of travel with less expensive actuators and smaller batteries. For this purpose, I do several things in my designs. I keep nearly all actuators as linear actuators concentrated near the center of mass. I use bicycle cables to reach out into the limbs. Further, I use electrical resistance to sense pressure, use gyroscopic stabilization and a RISC-V processor with vector extension. The gyroscopic stabilizer means the energy and complexity is for standing and walking or running and jumping is simple, robust, and minimal expense of energy. The centralized linear actuators allow for gearing via choices in threading and keeps unnecessary weight off the extremities of the limbs, meaning you can use less expensive motors and a smaller battery. The RISC-V processor with vector extension is also very inexpensive and draws very low power while providing significant local AI acceleration. Various options are readily available with Tensor Processing Units, Audio Processors, and other useful options. For example, some of the SiFive Mars V boards that even have Raspberry PI compatible I/O pins. Humanoids such as these could cost under $2K, potentially, in terms of parts and materials. With dual-axis gyroscopic stabilizer, it could even easily to backflips and other acrobatics and run over very rugged terrain as if it were second nature.

[@kampkrieger]

dual axis stabilizers? seriously?

[@solifugus]

@@kampkrieger yes. Look at cubli, for example. However, my approach is to keep both spinning and use pressure against them to move... This provides highly precise and strong control of orientation

[@rayr268]

I believe these robots will be better suited for war before house chores to get the development they need to tackle home tasks with outsized reliability

[@user-im5yj7uu4n]

100% this is why most car companies are getting into this industry through acquisitions and R&D. Perfect synergies there.

[@keenheat3335]

I also wonder about the precise actuator at a cost problem for a while. Though it was mainly for 3d printer space but it is applicable for robotic as well. Highly commoditized actuator that can move at nano meter precision does already exist. Your standard HDD driver actuator can move that accurate at extreme speed and acceleration at a commodity price. It achieve that precision via built in regular negative feedbackloop bldc motor with an end tip piezoelectric motor to control the finer movement. Though to achieve the nanometer precision, it does require thermal error detection and have the hdd disk provide position marking so the hdd arm can correct for error on the fly. But I think even without thermal correction, you could still maintain micron level precision at a very cheap price. How cost efficient to scale hdd arm actuator setup to lift human scale object and weight is hard to tell. But I'm assuming it's not going to cost too much from a cursory search from component list. The jump from commodity price to expensive luxury seems to occur at the robot arm level, you dont see such price jump for hdd actuator, if you compare from a % basis. I think this might just be a low volume assembly and packaging issue. Even that recent aloha robot, if you search for the individual component cost for the robot arm, it's like 30% of the cost. Personally I don't believe in the hype of bipedal human form factor, roomba with bigger wheel manage to navigate my house fine. Get some tank tread and suspension. Mobility is a pretty solvable issue. Roomba with a scara arm that has camera at the finger tip should probably be enough to solve grasp task. power wire management is a big ai issue though, technically there is an entire branch of math just to deal with how wire tangle with itself. So it's actually a diffcult problem. But if it can solve that, then power limit and battery life limitation is no longer a big issue. The robot can just have two wire and switch between the two when traveling room. Roomba suction pressure is like 1/10 of wire vacuum due the power and wire limitation, solving how robot plug into socket using a wire would solve a major hurdle. For outdoor robot that can use combustion engine to run a generator, the power and battery life limit is solved issue. The issue turn into combustion exhaust and noise, so it cant really be placed indoor. No where to empty that exhaust gas without harming human.

[@garethrobinson2275]

I like the hard drive actuator comparison, amazing precision. I don't agree that bi-pedal is the wrong format, though. What you are asking for is a disabled robot that cannot climb stairs, reach cupboards or jump in to a Robotaxi with you.

[@AleksandrVasilenko93]

We need better actuators AND better AI. Tesla Optimus already has the hardware to do many household chores, but its AI is not capable yet.

[@jeffg4686]

Regarding 15 minutes of charging for 45 minutes of use. that just means there needs to be one charging at all times. and then the robot just goes over and replaces it's own battery pack with a fresh one and keeps going. Just has to have a third, small battery that lasts long enough for the robot to replace it's main battery pack. All day long my friend. All day long (and night too since they don't sleep). One robot will replace 5-6 workers eventually because they can run 24/7, don't take breaks, and likely much less prone to errors than humans (eventually anyways). And they still can't see socialism yet (or won't admit to it). I wonder how their judgment will go?

[@TheJohnnyJohnny]

What if AI cam figure out how to design its own actuator based on real time knowledge? Are we humans doomed?

[@mikhailbulgakov1472]

Unitree has announced a very capable humanoid robot for $16,000. It will be available later this year. Because of protectionnist laws (to protect the wealth of billionnaires) it probably won't be available in the US but people living in Asia will be able to start living in the future with an humanoid robot at home. The future is already here for some.

[@garethrobinson2275]

We might not agree on what 'very capable' means but yeah, it is a shame that protectionism exists.

[@bobtarmac1828]

Swell robotics, however, Ai jobloss is the only thing I worry about anymore. Anyone else feel the same?

[@Dr.Kraig_Ren]

Westerners have a freaking easy life. Cuz we are rich, even low wage jobs can get you stuff like iPhone (you need kidney for that in other nations) Well, look at the developing world to see what'll happen after ai revolution. First adults will start replacing teens in part time jobs. And only one member of the family (most probably male) will sustain the entire family. In third world, stuff like vending machines and automatic ticket machines doesn't exist. They employ people. So automation has already "replaced" a lot of your jobs. West has a labour shortage. Migrants work for us.

[@yapdog]

In 2015, wrote a novel that touched on that very issue, where I predicted much of what's happening today and where it'll likely lead. I published it on 2020. No one read it, unfortunately. My guess is because it's a big literary/AI/psych-techno-thriller/PI story that's neither dystopian nor utopian. People only want stuff like The Creator which don't even say anything, but give a veneer of depth. ::shrugs::

[@HenryMaier]

Technology breakthroughs in one area motivate new technology elsewhere. For example, starlink internet would not be economically practical without reusable rockets.

[@tesseractgon-dy5yo]

why are their lower limbs slightly bent like they're preparing to poop

[@destinyglitches9519]

Robots can be useful without batteries. Corded robots

[@kampkrieger]

it is just wrong. yes the 'software' exists, at least in parts, the challenge is to integrate them in a reliable way and that is not trivial at all. the hardware on the other hand exists for decades. you couldn't have come up with a better example like self driving cars, where hardware is cheap and simple yet no company made the software work reliably despite the undisputed impact it would have on the market

[@CarboN428]

SOFTWARE (high level) was main bottleneck since latest LLM/multimodal AI, it just wasn't worth to manufacture advanced robots when they couldn't understand human environment.

[@ivando5372]

I am thinking robots repair shop.

[@liberty-matrix]

"Over the next 15 to 20 years humanoid robots will disrupt human labor throughout hundreds of Industries across every major sector of the global economy. The disruption of labor will be among the most profound transformations in human history." ~Tony Seba, 2024

[@ThomasTomiczek]

He is not wrong. I project Tesla doing 1 million robots in 2030 - per day.

[@FLORIDIANMILLIONAIRE]

That's very true, AI is a scam but actually robots are about transmission not the motor which has classically been same for over a century now.

[@ThomasTomiczek]

Ah, no. TONS of changes on the motoer side now actually.

[@FLORIDIANMILLIONAIRE]

@@ThomasTomiczek like what is the change in the fundamental physics of permanent magnets being rotated by electromagnets ?

[@ThomasTomiczek]

@@FLORIDIANMILLIONAIRE Let an AI answer. Remember, the claim was "the same for over a century". Here some changes in the last 50 years: Brushless DC Motors (BLDC): Introduced in the 1960s. Permanent Magnet Synchronous Motors (PMSM): Developed in the 1980s. Switched Reluctance Motors (SRM): Emerging in the 1980s. Axial Flux Motors: Gained attention starting in the 2000s. High-Speed Motors: Significant advancements in the 1990s and 2000s. Direct Drive Motors: Increased adoption since the 1990s. Integrated Motor Drives: Developed in the late 1990s and early 2000s. Silicon Carbide (SiC) and Gallium Nitride (GaN) Power Electronics: Became prominent in the 2010s. Advanced Cooling Techniques: Progressed notably in the 2000s and 2010s. Smart Motors and IoT Integration: Developed and widely adopted in the 2010s. Ups. Even the brushless motor is not a century old. Whow. What a claim. Hint: Ask an AI to invalidate your claims.

[@airliners321]

Wrong. Bad video. A very expensive and very useful robot will pay for itself. A cheap and dumb robot will never be in demand.

[@ThomasTomiczek]

Especially because a price of 20.000 USD (Tesla Optimus) may look high, but if you hit a 25 USD per hour minimum wage, that really breaks down to a simple decision in - months. less when you work shifts.

[@les_crow]

It doesn't require AI but you are talking about understanding context? lol.

[@TexasRiverRat31254]

Good, I need someone to clean my house since I can't tolerate modern woman enough to have one around. 😂🤣

[@ThomasTomiczek]

Sorry, this is crap. "why does it take so long to get robots into our households" and then going to actuators is dump beyond believe. Robots make no sense without the AI. The AI just showed up some years ago. The Tesla Optimus does everything inhouse and took what 1 to 1.5 years to where it is, max. That is not "so long", it is extremely fast. With all actuators made inhouse and developed from scratch. The AI was - and is - the real problem. Accuracy is good enough - heck, you do not even need that much accuracy because you can correct from the vision. Without the brain, no one puts money in the hardware. Only when the AI showed up, then suddenly everyone rushed. I really have a problem with "oh, it takes so long" when Tesla goes from nothing to a demonstrated decent contender in less than 2 years. This is not "so long", this is "whow". Now, the price - showing a total ignorance. Noone thinks there like NASA. 3 Prices are out. Tesla Optimus said to be around 20.000 USD - with a ROI of around 2.5 months in a factory. Ups. The Unitree is announced starting at 16000 USD - but options may take this higher and it is unclear whether i.e. hands are options. The 1X Neo - the CEO siad they target 8000 USD as price. None of them at 50.000 and all of them brutally cheaper than a human.