To Humanoid is to Err
Why humanoid robots don't make sense in the industrial world (currently)...
With apologies to Alexander Pope, who famously quoted “To Err is Human” I give you my 5 reasons why humanoid robots do not make sense in industrial settings.
Humanoid or “anthropomorphic” robots are the latest craze. At IREX (International Robot Exhibition) 2023 last week there were no shortage of new 2 legged and armed robots. While companies like Boston Dynamics have been working on humanoid robots for many years, there are dozens of new players in the “industrial humanoid” sector. Startups like Agility, Tesla, Ubtech, Figure, etc.. all have come out with bipedal robots aimed primarily at the industrial and more specifically at the logistics space.
The question is.. why? What problem do these designs solve? In this article I’ll address 5 reasons why I believe these specific designs of robots will not be a major factor in industrial applications anytime soon.
Let me address one point before we go farther. I know this article will feel like I’m picking on Agility Robotics but they are one of the few that actually has some real world video of their product and are applying them in practical industrial situations. They also have the most info out there currently. They created a bipedal humanoid robot that can move boxes around unaided. That is impressive technology. For those reasons alone they should be celebrated.
I have no issue with them personally. Many of the other humanoid robot manufacturers have nothing more than some cool marketing videos and promises of $20k robots in every home. I don’t give these companies much credit at all at this point. At least Agility has robots in the field.
My concern is that we are using this technology for the wrong reasons. Just like you don’t use a Ferrari to haul dirt you don’t race a dump truck at the track. Technology has uses cases and I just don’t necessarily agree with the current uses.
So anywhere you see “Agility” in this article you could just as easily replace with Tesla, APPTRONIK, Ubtech, Figure, or any of dozens of others manufacturers.
So again, nothing but respect for the people at Agility. I just question the application.
1. They Don’t Need Legs
Let’s consider the most famous of all robot duos, R2-D2 and C-3PO (or for your mega nerds, Artoo-Detoo and See-Threepio). C-3PO is your classic humanoid robot with two arms, two legs, a head, all the basic humanoid parts… Meanwhile R2 is basically a cylinder on wheels. Most would argue that having bipedal movement would be a huge advantage, but that depends on the environment. When the pair get stranded on Tatooine it’s obvious that 3PO has the advantage on the rocky and sandy terrain. Poor R2 even gets stuck in the stand at one point.
So in the real world, where we have rocks, stairs, uneven floors etc.. clearly bipedal robots are going to have an advantage. Look at Boston Dynamic’s videos of Atlas doing parkour. It’s not only amazing but possibly a bit frightening.
But industrial settings are not the “real world”. Factories are designed to have flat floors like the Tantive IV or the Death Star in “A New Hope” (OK, yes I’m a Star Wars nerd).
You don’t design a process that requires humans to walk up stairs to do their repetitive task. You don’t have random obstacles in the paths of people. Factories, like Rebel starships or “fully armed and operational battlestations”, are designed with flat floors with as few obstacles as possible.
So why do you need legs? Legs have no less than 2 axes of motion each (knees, hips) and many of these robots have more (add in ankles). That’s a minimum of 4-6 motors to control this motion. One humanoid robots manufacturer states that it has 46 axes of motion in the robot.
Furthermore legs are inherently unstable. The human body is CONSTANTLY adjusting it’s center of gravity to remain standing. Robots are no different. Gyroscopes and small tweaks to the motors are required to keep the robot from falling over. This consumes energy and requires extra processing power.
If the robot however is on wheels or tracks this problem is eliminated. Not only do you need less power to move the robot and have less motors, in the case of a power failure (like the robot above) the robot simply stops; it does not crash.
See that floor. It’s flat. Like 99.99% of all factories. Why add complexity and points of failure if they are not needed?
This is BellaBot from Pudu. (R2-D2’s distant cousin?) Intended primarily for restaurants (or hospitals etc.) it happily moves around on wheels delivering items to the intended location. It serves basically the same function as a bipedal robot with far less complexity. While this sector has also struggled I feel their strategy has a much better chance of success.
(Free startup idea: Put two arms and a torso on SPOT and call the company Centaur Robotics. The VCs will line up to give you money. I get 5% equity for the idea. You’re welcome).
2. They are slow
Bipedal robots are slow. Maybe even slower than cobots (and that’s slow).
Notice how slow these robots are to move these totes to and from the conveyor. And now notice (watch the humans) that this video is actually SPED UP!
There are a dozen ways to to this same operations with not only industrial robots but even hard tooled automation. It seems to be an unnecessary complication in an otherwise simple operation.
To be fair to Agility the Teslabot is no faster.
Nor is this robot featured at #IREX2023 (click to see tweet and video)
3. They are Expensive
While I’m sure the price point has come down a bit, according to this April 2023 article the price for a single Agility Digit robot is $250,000. Let’s say they even knocked 25% off the price. That’s still $187,500 for a single robot. To keep up with the lines at Amazon shown above that’s $375,000 for two robots. And based on battery life and charge times you’re likely to need at least 6 (and maybe more) robots to keep this line running 24/7 (see more below).
And let’s take a moment to address the others guys making outrageous claims of $25-45k robots in homes. Looking at you Elon.
If Digit is in the low 6 figures there is NO WAY anyone is going to be selling these things for $25-45k and making any money anytime soon.
4. They Rely on Batteries
Mobile robots have an inherent disadvantage to their stationary counterparts. They run on batteries. According to this article the battery life for Digit is 1.5-3 hours depending on the application. So assuming in the above scenario where you need to moves totes with 2 robots at 3 hour run times (and assuming a 3 hour charge time?) that means you need a minimum of 16 robots (24 hours / 6 hours per run/charge cycle * 2 robots = 8) to run this line 24/7.
At $187k a robot that’s almost $1.5M to move totes around (using optimal numbers).
I found a video of a previous version of Digit that requires a human to plug them into charge. In reality they need to be able to walk over to a charging station and recharge automatically (and maybe they do already..?) But 1.5-3 hours is nowhere near long enough to be effective. Heck, humans work longer without breaks. Until we can break a 8 hour shift on one charge they will continue to be harder to adopt.
5. Safety is a Unknown At this Point
As there are, to my knowledge, no current rules, regulations or standards regarding humanoid robots at this point (not sure if ANSI/A3 R15.08-2-2023 covers this class of robots), it’s the wild west out there. A recent article about Agility states…
Due to safety regulations implemented by Agility Robotics, humans must also stay six feet away from Digit. “We need thousands of hours of robots operating to show and prove the safety case for those before we feel comfortable,” said Hurst. He estimates it could be several years before the robots could move and work seamlessly alongside humans.
So we should applaud Agility for taking the safe route by keeping humans away from the robots. But how? Does Digit STOP when a human approaches? Or do we need physical barriers to prevent humans from entering the area?
What’s the plan moving forward? Force limiting? Laser scanning? What good is a humanoid robot if it has to remain in a pen?
The entire point of humanoid robots are for them to work alongside humans seamlessly. But without a proper plan for safety this is never going to materialize.
Conclusion
So while companies like Agility, Boston Dynamics and Tesla have gone all in on humanoid robots and the technology is AMAZING, I just do not see the use case in a controlled setting like a factory or office or hospital etc, where obstacles and stairs are rare. While they will get better, they are currently slower, more expensive and less efficient than their industrial counterparts. The unknow safety aspect also limits the usefulness of these bipedal robots. I just see a lot of headwings to scalability when there are already solutions that solve the problems.
These technologies will get better and will get cheaper. But I do not think the industrial world is where they should be putting their efforts.
The two spaces that I do see these robots being successful in the short terms are (unfortunately) Warfare/Policework and Search & Rescue/Firefighting. Even here quadrupeds like SPOT are probably better suited.
So in closing, if I’m on Tatooine I’m picking C-3PO. If anywhere with nice flat floors (like 99% of industrial settings) I’m team R2-D2.
The Automation Navigator is brought to you by Automation AMA, a company founded and run by an industry leader in assembly, robotics and packaging. Automation AMA offers a variety of services to help you on your automation journey or to help fine tune your engineering, sales, operations and business development departments. Learn more at www.AutomationAMA.com or shoot me a introductory email.
Good article, Sean! I love the story. Agility started up in my hometown with engineers from Bill Smart's robot engineering program at Oregon State. Humanoid robots are the rage because BD finally proved how they can work. MIT / Vecna developed "BEAR" 20 years ago (Daniel Theobald's company). Mankind always wanted a look-alike robot so we can feel God-like (Mary Shelley's Frankenstein was an early example; Gollum from the Dark Ages, which became "orcs" in Tolkien's writing is another). Now we can finally do that. And so a millennia old thirst is quenched.