Honda calls the machine it revealed at CONEXPO-CON AGG 2023 (that’s the largest construction trade show in North America, and, yes, that includes earthmoving and aggregates, so you know it’s for real) an Autonomous Work Vehicle (AWV), which looks a bit like a friendly, cab-less, flat-bed pickup truck. Honda calls this a vehicle, but, really, this is a robot — a robot designed primarily to haul and deliver things, which is why it looks the way it does. The AWV can operate autonomously in a wide variety of industrial and construction environments, and can be manually operated by remote control if desired. Seeing a robot like this I think is a good reminder that despite the hype that companies like Boston Dynamics and Tesla and even Honda itself have gotten for their humanoid robots, it’s important to remember that for so many jobs, something shaped like your uncle isn’t necessarily the solution. Robots can be like animals, with their morphology dictated by the job they need to do, and this Honda AWV is a perfect example of that.
Here’s a video of the AWV in action:
It’s worth noting that this is actually the third iteration of the AWS, with earlier generations already having undergone a lot of testing, as you can see in his video from 2021 showing a lot of second generation AWS robots working on a construction site:
Interestingly, it appears that a lot of the parts for the second-gen AWS are coming from Honda’s Kei-class pickup trucks, as you can see here:
The taillights and most of the sheet metal of the bed, including the tailgate and, um, sidegates, all seem to come from the Honda Acty, as you can see up there. Those are all extremely proven components, and I wonder how much of the frame and suspension are borrowed from the Acty, too.
This new third generation no longer appears to be built from Kei-truck parts, and now I wonder how much of the drivetrain is shared with the Honda e city car. Honda also seems to be accepting of and relying on how people naturally tend to anthropomorphize machines, so the AAWV has an obvious and friendly “face” which might help the people who work with the machine interact with it in a more forgiving and comfortable manner:
The third generation AWV is updated from the previous version in a number of ways, most of which were based on what Honda learned from their field testing of the second generation at a large-scale solar panel construction site in the Southwest. The press release describes the changes for the new version:
Key features of the third-generation Honda AWV include:
Increased bed size to 2 pallets and greater loading capacity of 2000 pounds
Improved navigation in locations where GNSS (global navigation satellite system) service is weak or unavailable through the use of LiDAR sensors
Simplified tablet-based programming interface and cloud connectivity
Higher speed in autonomous mode – up to 10 miles per hour (mph)
Increased battery size and longer operating time of up to 10 hours
Enhanced avoidance function for vehicles stopped on road
Lower bed for easier loading and improved ergonomics
The use of LiDAR is interesting here as a method to compensate for poor GPS signal navigation, and I’d be curious to get a better look at that “tablet-based programming interface.”
The AWS as configured is primarily for jobs that involve the deliver of materials and supplies – up to 2000 pounds on its eight foot long by over four foot wide bed – to precise points along a given route, a common and near-constant job for many construction-type projects. The same fundamental AI, sensors, and drivetrain design could likely be reconfigured for other physical tasks beyond supply and material delivery.
I do think our future will include more robots, though I suspect that the ones we’ll see will be designed to execute fairly specific tasks as opposed to a general-to-everything-use humanoid-shaped robot. I know it may seem like I’m being harsh on human-shaped robots here, but think about it: at the moment, there are close to zero humanoid robots actually in actual, day-to-day, get-shit-done industrial or commercial use. I’m not saying that one day there won’t be a place for them, but at the moment and into the future, if you need a specific job done, a robot that is designed to do that specific job tends to make more sense.
In short, more R2-D2s, less C-3POs.
I do think there is a future for humanoid robots in the medical field, as getting a diagnosis from something that at least LOOKS human would probably be easier. But outside of that application I can’t see a case where a humanoid robot would be better at anything.
Paging Austin Powers…
I think these things are bloody brilliant! Robots like these have been kicking ass in warehouse settings for ages this is a very natural evolution. Well defined functionality in a defined environment is an excellent use case for self driving.
But Amazon will never use them, because Bezos can’t force a robot to pee in a bottle
As I was looking at these videos, I was wondering about those dusty environments? Don’t LIDAR and cameras suffer the most in those environments? I feel like GPS is where that would do the most good, because of the limited vision and the LIDAR scatter from dust.
But what if your uncle is shaped like a fork lift?
It’s interesting that feedback from their solar test project led them to increase the bed size to two pallets, but only raise the payload capacity to 2,000 lbs. A pallet of solar panels weighs 1,500+ lbs, so in the application where they tested the previous generation, it’s still only a one-pallet machine despite the increased bed size.
I’d suggest the solar industry employ Bigfeet until Honda comes up with an even higher-capacity model, but a few inches of the second pallet would be hanging off the back of one of those…
Agree that most robots will be distinctly nonhumanoid, but you have to admit Tesla’s Technoking/CEO robot is eerily lifelike.
I don’t really think it’s more lifelike than the one Facebook has. They just dialed up the antagonism, which makes it seem more interactive.
I heard that they trained the earlier version by just letting it loose on Twitter and within hours it had learned to be a huge racist.
What a change to have it tweaking the Twitter algorithm to promote racism instead. We really have come a long way.
I’m thinking these could be successful elsewhere, but in the USA they’d be mocked as socialist unless they are F-150 sized with evil faces, and randomly hunt down and kill 1% of workers to prove how tough they are.
Tesla’s approach to automation is indeed similar to the way a 12 year old would approach the problem, and not only for their humanoid robot. Same goes for their self driving project. Humans use their eyes to navigate their environment, hence cars need to have “eyes”! Disregarding that perhaps “eyes” were never the best tool for that task, just the one we happen to be equipped with…
So Roombas, not Rosies. Man, The Jetsons was wrong about everything…
Just sent our robo-vac to go sulk on its charger again, we’re still trying it out and so far I’d rather go back to using a broom a couple times a week.
I just got a robo-vac too, although in the interest of full disclosure mine was pretty cheap so it’s probably not state-of-the-art. It’s funny because sometimes it seems really smart, and then the next instant it will do something really stupid like repeatedly run into the back of a chair until it gets itself wedged in and can’t escape.
Add in the fact that the filter seems to clog really fast and it has no ability to define off-limits areas (so there can’t be a single cord or anything like that on the floor anywhere that it can reach) and I’m also not sure it’s worth the hassle. I certainly wouldn’t trust it to run on a schedule when I’m not around to keep an eye on it.
They were right about soul-crushing repetitive work being something we just don’t want to get away from, even if we can automate everything.
Sure, it’s not the part any of us hoped for, but they weren’t all wrong.
Humans are perfectly shaped for approximately zero jobs, and the idea that humanoid workers will be wandering around factories or job sites doing human jobs *in human ways* is absurd. It’s “wouldn’t it be cool if …” thinking in place of “what would be the ideal way of executing this task/solving this problem” thinking. There is no universe where Teslabots are going to move stacks of drywall around better than a robo-flatbed.
Our morphology is such that we can’t even be “stored away” without the need for an external device (chair, bed…).
But by all means, let’s make our very heavy, fragile and expensive robots as tippy as we are.
Reminds me of Gallagher asking why both our feet point in the same direction.