A guy named Ken Fox, who posts frequently at X, designs automation systems–electrical components, pneumatics, and software–especially for the food processing industry. Here’s his ice cream cone filler at work: video.
There are a lot more videos at his X feed.
When people talk about manufacturing, they usually seem to think about metalworking in one form or another–but there are other important types of manufacturing, including the process industries…refining, fertilizer manufacturing, plastics processing…pharmaceuticals manufacturing…and food processing.
Also, I notice that a lot of people judge the level of automation in a particular company or across an entire national economy by counting robots. I don’t think this is a very good metric. How many humanoid robots would it take to equal the performance of Ken’s ice cream cone filler, or any of the other automation systems in his video collection? You could in principle make a CNC machine tool by having a humanoid robot turn the wheels on a manual machine tool, but it makes a lot more sense to just mount the servos directly on the machine. Similarly, elevators could in principle have been automated by having a humanoid robot handle the controls, but it was simpler to just build the logic into the system.
There will be a big role for humanoid robots, certainly, but I suspect that in many cases they will be a temporary bridge to a more comprehensive system.
Anyhow, enjoy the videos!
Well, there’s generic machine tools, and there’s “transfer machines”. Either one can be CNC, or not. The used to frequently run on cams. They’re usually CNC now, because the technology is available.
In a generic machine tool, a part is loaded, and toolholders are loaded and changed until the operation is concluded.
In a transfer machine, the part is loaded in a fixture mounted on a transferable base called a pallet, and when the specific stage is done, the pallet is transfered to the next machine for the next step, sort of like a machining assembly line. The lines are usually rigged in ‘U’ shapes in plant, rough parts loaded on one side, and finished parts come off the line 50 ft away.
It how engine blocks, etc are produced in a production plant. Generic full use CNC machines are used to make prototypes, custom work, and rework repair. That’s not how you make parts when the quantities run into the millions.
The ice cream filler you showed is a transfer machine.
Can you tell I worked in manufacturing way too long?
A humanoid robot running a lathe or milling machine wouldn’t be much faster or more accurate than a human doing it. It would be horribly inefficient compared to a normal CNC controller. The servos controlling the ball screws that move the machine axes are almost infinitely less complicated than an anthropomorphic hand capable of turning a dial, connected to an anthropomorphic wrist, forearm, etc. The brain of a CNC is pretty simple as such things go, far less powerful than the simplest smart phone. I doubt the robot could control 2, 3, 4, 5 or more axes simultaneously with a resolution in millionths of an inch to produce contours that are all but impossible on standard machines. And do it at speeds that may be 10x faster than possible on a manual machine. A humanoid robot would require orders of magnitude more processing power and complexity just to stand up and walk to the machine.
The sorts of machines in the videos have been around for a long time, a hundred years or a bit more. For most of that time, they were controlled by cams, gears, links, etc. They were also fiendishly difficult to keep running. There are two big advantages to the new designs. The first is the user interface that he’s rightly proud of. If an adjustment is necessary, changing a value on the screen replaces endless fiddling with wrenches and feeler gauges or “quarter turn” tweaks here and there, with one adjustment possibly affecting several others. Also as shown, when something goes wrong, it can direct the operator to the likely culprit which isn’t necessarily obvious. The second is the mechanisms themselves are much simpler, cheaper, more robust and less wear prone than the collection of cams, levers and rollers that they replace.
ed…yes, I understand that point. I wasn’t arguing that a machine tool, CNC or manual, is the same thing as a transfer machine. My point is that there are lots of kinds of productivity-improving technologies of which humanoid robots are only one. Maybe a transfer machine has a computer vision system added to perform inspection functions, or a mechanical arm to add components to an assembly. Seems to me that the full functionality of a humanoid robot…ability to move on its own, ability to work closely next to humans without risk to the humans, etc, is only really needed in certain situations.
“Thought you guys might like this ice cream cone filler I designed the automation for. I designed the electrical enclosures; selected all electrical and pneumatic components; wrote the schematics; hand built all the electrical panels; then I wrote all the controller software and UI software…”
Because this SUPERIOR to going to your neighborhood ice cream shop?
You clowns should find something better to do.
Because this SUPERIOR to going to your neighborhood ice cream shop?
You clowns should find something better to do.
And such small portions!
I won’t be as nasty as anon … but I am scratching my head a bit at the focus in the post on humanoid robots. As ed and MCS point out, nobody did it that way, and probably nobody will in the future, either. In theory offices could have automated document distribution by having thousands of robot carts moving physical paper around (a variation on the pneumatic tubes sometimes used) but everybody realized that digitizing documents/information and sending links was far more efficient, even going back to the days when it was dumb terminals rather than PCs on people’s desks. It seems unlike that humanoid or animatronic robots are going to be anything but a niche for special applications like mobility over highly variable terrain. Anything automated in a controlled environment is going to be designed to function specifically in that environment, and most general use robots will be designed for specific tasks (like a Roomba) or operate to assist humans rather than replace them.
Christopher B…”but I am scratching my head a bit at the focus in the post on humanoid robots”…Elon Musk, among others, has been talking up humanoid robots a lot, and I’ve seen a lot of posts & articles using penetration of such robots as an indicator of technological advancement and productivity. For example, a recent Noahpinion post on manufacturing productivity”
https://www.noahpinion.blog/p/why-has-us-manufacturing-productivity
which is being discussed at LinkedIn:
https://www.linkedin.com/feed/update/urn:li:activity:7221559334810398720/
includes this quote:
“In 2021, China had installed 18 percent more robots per manufacturing worker than the United States. And when controlling for the fact that Chinese manufacturing wages were significantly lower than U.S. wages, China had 12 times the rate of robot use in manufacturing than the United States…The United States had 274 robots per 10,000 workers, while China had 322.”
It’s one metric, but IMO not a good proxy for overall technology level in manufacturing.
My wife and I have reached the point where yard work (on multiple acres of land) is getting difficult, and young teenagers willing or able to do such are in short supply. (As is the quantity of young teenagers, period. We have strange demographics in $TINY_RURAL_TOWN.)
It doesn’t have to be humanoid, but a general purpose robot that could find pine cones or weeds and such issues would be quite handy. Alas, my engineering background dealt more with manufacturing integrated circuits, rather than the robots that might use them.
Paging Daniel Boone Davis: Flexible Frank has a job to do. h/t The Door Into Summer
A robot to simple lawn chores would and probably does look like a ridding mower without a seat, maybe with a vacuum to pick up pine cones and a spot sprayer, On wheels, not feet and legs. Deere is close to autonomous tractors and combines and sprayers that use machine vision to spray just weeds rather than areas are getting sort of common. They look just like regular machines because they are. Navigating a premapped field is a lot easier than self driving but still isn’t there yet. Then there’s the issue of layering more complexity on top of very complex machines and the negative effect that has on reliability.
Back to the lawn machine, In principal it might be easier, but not much cheaper. It’s easier to justify the cost on a $750,000 tractor or $million combine, not so easy on a lawn mower. Now if you want one that can climb trees to trim them as well, that’s going to take a while.
As far as counting robotic noses, any statistic out of China is just not worth paying attention to, especially if it’s being used to demonstrate the superiority of China over the rest of the world. As anyone that has followed the U.S. Commerce Dpt. and BLS should realize, the only question is just how wrong their numbers are, in which direction and what agenda are they trying to advance. Comparing Chinese fantasy with American fantasy is a fool’s errand. On top of that, there’s the whole question of just what qualifies as a robot anyway.
“In 2021, China had installed 18 percent more robots per manufacturing worker than the United States.”
In the great scheme of things, that does not sound like so much of a difference. Further, it is a gut cinch certainty that most of those robots are custom-designed machines to perform a specific repetitive task quickly & efficiently. Just look at any of the many YouTube robot-rich videos of automobile manufacturing plants in China. Those are robots — but they are not humanoid robots.
For general purpose assistance — wash the car, clean the gutters, do the dishes, take care of grandpa — it is hard to see how a humanoid robot could be cost-competitive with a Filipina maid. And Filipinas have nicer smiles than any humanoid robot.
A guy named Ken Fox, who posts frequently at X, designs automation systems–electrical components, pneumatics, and software–especially for the food processing industry. Here’s his ice cream cone filler at work:
The father of a lifelong family friend invented a process for manufacturing ice cream cones. For at least a century the company was run by a family member.
The robots most people think of are the arm type, available in all sizes from smaller than human to much larger, capable of exerting thousands of pounds of force. They’re fairly simple by themselves, capable of following a programed path with precision, not that different than a lathe or milling machine. In fact, it’s possible to attach a machine spindle to the wrist and do machining with it. Many do little more than pick up an object at point a and deposit it at point b. It’s when they are coupled to something like a machine vision system that they become capable of more complex behavior but it’s the vision system or other sensors that calculate the path for it to follow and feed it, usually in the G-Code used for CNC machine tools.
The 18% robot gap sounds like the windup of some politician promoting a new “Moon Shot” sort of program where enormous amounts of money will be channeled to preferred constituencies without changing anything but bank balances.
I remember as a kid watching the opening scene of “Terminator 2”, where the Resistance was fighting the Skynet army of robots, and wondering why there was anthropomorphic robot infantry. Well kids understanding the art of movie making and all of that
I’m sure people pushing humanoid robots aren’t just thinking about their utility.
Noah Smith had an interesting point in that Substack regarding that one reason for low American investment in capital equipment is low wages. A valuable concept I learned a while back is definition of “technology” expanded to mean any ability to deliver a certain outcome. For K-12 education it could be home schooling, distance ed or computer-based training,. charter schools, or the public school system. For business there is more than one way to be in manufacturing, depending on what business you are in and what you want to achieve. In a previous post, someone mentioned that early textile manufacturers should have just bypassed machines and just shipped their industries to labor-intensive India; both are example of different technologies.
We have been dealing with the ramifications of technological investment in manufacturing with the Japanese since the 1970s with the Toyota-model of manufacturing, just-in processes, and robotics. One of the things we have learned and that my guess is the Chinese will learn is that all that capital investment might look good in the papers and win you plaudits with the productivity-trolls, but it better pay off on the balance sheet
To get back to the ice cream machines for a moment. It would be perfectly admissible to produce the ice cream cones by using a robot arm to pickup the cone, hold it under a spraying station and then, possibly, use another robot to place a scoop of ice cream. This would count as high tech robotic production, it would, maybe, make 1/20th as many cones in an hour as the dedicated machine and we haven’t done any packaging yet. The robot would let you more easily make individual cones in different sizes, with different flavors, etc. But that’s not what’s wanted. You could also unbolt the robot and put it to making auto parts, not so the dedicated machine. The proper level of automation and robotization is whatever gets the job don cheapest and the entity least likely to get that right is the government with politicians picking winners and losers.
Mike…”Noah Smith had an interesting point in that Substack regarding that one reason for low American investment in capital equipment is low wages.”
Certainly not low wages compared with China, even though wages have risen there…definitely not compared with Vietnam or Indonesia or any of several South American offshoring destinations. It is certainly true that higher wages drives more productivity investment, there will be investments that make sense if you have to pay $25/hour that make no sense at $10/hour…high wages have historically been a driving force for mechanization in the US. If it’s not working now, I have to think that there are factors other than wages-not-high-enough that are in play.
Comes now Megan McArdle, arguing that we cannot return to more American products being made in the USA because consumers won’t accept the higher costs.
https://www.washingtonpost.com/opinions/2024/07/25/vance-economics-politics/
Yet tariffs were fairly high during much of American history, during which time American wages were high and growing by world standards.
If refrigerators absolutely *had* to be made in the US, for whatever reason, then the productivity of refrigerator production would be driven up by competition among domestic producers. Refrigerators per man-hour is not a constant.
US factory construction investment over time.
https://x.com/erikbryn/status/1816625773555699962
Some of this is surely in response to government subsidies, many of which will prove to be misdirected..but I doubt that all of it is.
David, agreed regarding the wage differential. Noah actually hits on one of the key points regarding comparisons between China and US investments and that is “Of course, Chinese manufacturing wages are also low, but the government puts its thumb very heavily on the scale there, encouraging capital investment”
The key to investment is return but what return exactly is the Chinese government seeking by pushing hard for large-scale capital investment? Perhaps a different return (financial) than American companies, one more political. There’s a different game afoot, with differing priorities, certainly not an apples-to-apples comparison. China is somewhat infamous for encouraging over-investment for certain sectors leading to disastrous results.
That doesn’t resolve the issues regarding investment in American manufacturing, but the Chinese economy marches to a somewhat different drummer than ours.