Most robots walk like they've soiled themselves; this one walks with a crazy amount of personality. Disney is getting very good at combining
By Loz Blain, October 08, 2023
"Disney is getting very good at combining the art of character animation with the science of bipedal robotics, and the results are … adorable."
"Unveiled at the IEEE IROS conference in Detroit last week, this new, unnamed robot is clearly based on the BD-1 designed for 2019's Star Wars Jedi: Fallen Order video game, developed by Respawn Entertainment."
"In order to make him look more friendly and appealing, BD-1 was designed to mimic some dog-like behaviors. When he's curious about something, he'll tilt his head. His antennae are able to rotate, and they mimic the movements of a dog's ears. He comes across as earnest, inquisitive and downright adorable."
The robotics world is buzzing after a demonstration by Chinese firm LimX Dynamics showcasing the remarkable agility of their recently unveiled Tron 1 robot. Initially launched with a focus on modularity – allowing it to transition between walking and rolling locomotion – the Tron 1 has now proven its capabilities far beyond simple traversal, exhibiting impressive acrobatic skills that rival those…
ROBOTERA Unveils ROBOT L7: A Full-Size Bipedal Humanoid Robot That Dances and Works
Beijing, China — (ARAB NEWSWIRE) — Recently, robotics company ROBOTERA officially launched its next-generation high-performance humanoid robot—the ROBOT L7. Standing at 171 cm tall, this bipedal robot is capable of high-dynamic actions such as running, 360-degree spin jumps, and street dancing, while also performing functional tasks in logistics and industrial settings like sorting, barcode…
Part 2 of 2: The evolution of the walking moving thing called "mecha"
So in the last post which you kind of need to read and look over the videos for for this to make sense, we went over wheel strategies for something which functionally was akin to a snake learning to become a rodent or a thing with the beginnings of very simple limbs, and we saw a machine that could tip up onto wheeled feet to balance.
Gunhed is something of an overdesigned example, but we are seeing tangible benefits: We have the beginnings of a vehicle that can cross terrain pretty much no conventional road vehicle can.
I think what its missing here is a long scissored foot, which can close around right angles, to create grip via compression, as seen here.
I would embed, but that hits the post limit.
But the next stage is to think about going upright. Up on two legs. Not necessarily walking, but going upright and making this useful.
But why would we do this?
Well, Boston Dynamics' Handle is a great example. I personally think this is far far more interesting than Atlas, because its movement potential is actually far far closer to how humans move than Atlas.
But it doesn't even walk, its not even person-shaped!
Let me explain:
If you look at a lot of walking robots, you see the same thing again and again, and again:
They're technically not even walking. This is called dynamic repositioning, but it technically isn't locomotion. Why?
It doesn't balance its hip, or its upper weight as it moves.
Why is this important?
Well, in ancient Guinea in West Africa, some some six to three million years ago, early humans combined ape and human like ways of movement. The initial motive was being taller made you more frightening to predators or when competing for mates, and it gave you a better vantage view when searching the horizon for threatening silouettes or shapes as Meerkats do.
They used social signalling to help call out threats early giving time to formulate a response -- which we probably started doing somewhere between 2 million to 50,000 years ago depending on your vantage point.
Being upright enhanced this capability...
But why is it important for locomotion?
First, locomotion - real locomotion - is all about using gravity to pull you toward the ground, and then catching yourself with your feet like a unicyclist who then makes gravity do a whole bunch of the work for them.
Note the obscene acceleration and variability in the movements of the riders here. This guy truly is making gravity his bitch.
What advantage did this give us? We went from being opportunistic hunters and pursuit hunters who go extremely fast to initially and very briefly becoming persistence hunters who would basically follow stuff to death not by running, but by travelling huge distance over long periods of time by sweating and being more efficient than four legged animals.
This did not last long, however, because the other advantage we gained was far more interesting:
Seeing objects from further away gave us the need to become complex planners, who could estimate and spot threats due to the new wealth of information available to us and quickly we dropped persistence hunting for planned hunting but also long-distance navigation and navigation planning.
Second of all, the other great advantage? We could scale other terrain other animals couldn't easily, efficiently which allowed us to spread our territories much further and to collect resources and relocate ourselves.
This combined capacity for picking ideal areas and having excellent pattern recognition meant we were able to pick out the most ideal targets -- the weakest of a given group.
What about hard landings?
If this could be scaled up, they make even the meatiest carrier landing gear on fighters look like peanuts.
The Tomcat here is a very heavy relative to modern fighters, and look at how the gear just eats up that landing. If the deck weren't flat, it wouldn't do anything of the sort.
So what about a foot? We mentioned a foot previously!
I think of stuff like this. See how its able to use the "foot" to assist its climbing, even on a surface which has no opportunity for grip otherwise?
If this was combined with the robot design in part 1, lifting up onto its front, you conceivably have a design which can then use this front equipment to capture the angular stress, essentially becoming a foot if it opens sufficiently.
It also then means if the back wheels are raised behind the design, and the platform is kept level and the same reaction-wheel like principle seen here in the AGRO2 concept in the previous post, while having a more durable form of the morphing track wheel we saw, we start to see the capacity for something which really can move dynamically.
Its not walking yet, but it is skating, and it is using all of the shock absorbsion and standing principles of a walking machine -- and may potentially with further advancement of that knee design become a walking machine.
Or maybe we do something entirely different? such as tracked skis? (I'm hitting the video limit again, sorry)
Okay, but what if we have active shock-absorbing in the foot?
The limitation here is his hips aren't wide enough to swing properly -- since a human spine acts as a sort of force distributor. If you've ever seen a shake go up through a spring, this is why you see robots without functioning hips when they take steps really struggle. Remember Asimo? Walking like it shit itself.
Verses Petman, which has a proper hip, and a simulation of something roughly resembling a spine?
Can't embed it easily due to the video limit, sorry.
The floor compressions are distributed mechanically, rather than actively through some sort of simulation which has to guess what the impact from the ground is going to send up through the body, to reduce oscillations.
Though this actually a problem Disney sort of solved, by using wholistic full body balancing.
It doesn't fit here in all of this, due to the embed limit, but here's the video. Its genuinely VERY impressive.
But what about active movement, or active balance compensation and support? What happens if we go beyond the form?
What about propulsion?
Meet LEO. Its using thrusters of sorts not for full flight, but to provide thrust to offset and provide the extra nimble action it needs to make finer movements.
You can sort of think of it as being like a person with a jet engine who's not using enough thrust to fly, but enough to make a very high jump to fight SOME of gravity with all of this dynamic positioning.
But what about flight?
What about active movement in flight?
Well, we see that with this:
Sure, its heavy for now, and its not very fuel efficient (packing just 10 minutes of flight time, at a thrust/weight ratio of 1/1) but this is assuming we're going with a petrochemical engine or that we don't use to some sort of alternative ultra high density power-source.
Yeah that's basically cartoon science fiction right now with how limited the energy density of batteries, but in the next forty years it won't be especially as we see movement towards massless batteries where mechanical supporting structures themselves become the battery.
But this thing is woefully inefficient with energy. Its HOVERING. Its generating all of its own lift instead of getting the atmosphere to do it! That's SO energy efficient because he's spending thrust to do everything!
But this is so inefficient! We could be using a wing!
Sure.. But this guy isn't!
So what happens when we start putting all of this together?
We honestly don't know yet. Its never been done.
But combinations have been done on smaller scales.
Okay, so what happens when you combine say, an engine like this with wheels at a lower output?
Quite a lot, actually in terms of both agility and speed.
Whenever these features are added, we start seeing moments where one technology multiplies the other's potential functionality, capacity speed and even energy savings.
And this is the thing.
When you look at this thing we call "giant robot" and its many parts in isolation, its all total nonsense jibberish. None of it makes any sense.
But if you make it smaller, and then you consider how the different functions amplify one another, you start to run into something very interesting:
They make an uncomfortable amount of sense.
See this is how invention usually happens.
How such an evolutionary path may function.
Not with truly synthesizing a new idea out of thin air, but to quote Raph Koster and his excellent talk on practical creativity for game designers, you combine things in a way which transforms the context to create a new usecase, or you solve an old one.
We're not even talking about arms for balance or firecontrol like some kind of insane modular Close In Weapon System system capable of shooting down incoming ordanance precise non-line of sight fire yet, or even any of the other limb potentials like performing logistics or construction to establish forward operating areas, and yet we already have in our minds something which is hypothetically useful.
If you're interested in more, here's an INCREDIBLE video essay series by my friend Argonbolt which goes into not only mecha as a functioning thing, but also mecha as a fictional medium and the future of that genre itself. If you like giant robots in any capacity, you NEED to see it.
Take care, and I look forward to any replies.
Edit: Future essays will be collected here. I'm currently collating past rants. My next is on the nature of why transformation is good, actually. Hope you you enjoy it!
Part 1 of 2: The evolution of the walking moving thing called "mecha"
Let's assume for a moment, that we deal with mecha functionally first, rather than aesthetically first.
What are those functions?
Well, I wrote a list nearly two years ago which should give us a handy start:
Osaka's "what it means to be a mecha"
Last updated (15/01/21)
Subject to change, via ongoing discussion and refinement in communicative terminology.
This communicates the terms of mecha in the sense of a pure vehicle, rather than the artistic expression or cultural notion of "mecha" which I believe are very different things from what is described here.
1. The system must be designed around omni-directional movement and be able to fire at objects which are not in boresight
2. The primary equipment mounts must support movement in some way, such as varying the centre-mass of the body via posture or allowing a platform to right itself if it falls. Secondary weapon mounts do not need to meet this requirement.
3. The platform must be able to climb to platforms and ledges which have an interior gap at the point of classical accordance which is greater than one length of the platform's body
4. The platform must support the dynamic repositioning of its centre mass through large moving elements
5. The fire control system must be capable of both firing based on cued tracking information, or along a given vector relative to the platform's forward 180 degrees. It should also be equally effective against ground or air targets via sensor fusion and situational data synthesis through the use of situational intelligence software systems.
6. The platform must be able to endure a landing at 30% of its terminal velocity and be able to continue mission duties conventionally
7. Although not a must, the platform should support a variety of non-combat duties such as engineering, construction and maintenance tasks essential in the establishment of a forward operating base with only minor changes to its loadout
Okay, well, let's start on 3, 4 and 6.
They are afterall, the hard problems.
Okay, so where are we right now, and how did we get there?
So DARPA had a really bright idea recently, which was to work on a platform which is more about evasion of objects on the ground and fluent ground motion than it is using sheer mass to tank or withstand a blow: Ground Vehicle-X Technologies:
Don't get me wrong, it looks a touch silly here, but when you think about the actual capabilities of the form, if you know some vehicular design principles it kind of makes sense.
The beginnings of this thing are 4x4 extreme hill climbing vehicles in the 1980's, which has been a sport for many many decades now, using ultra lightweight material and insanely high performance engines that aren't really designed to last built around turbocharged four stroke twin turgocharged V8 engines which did something really clever:
They weren't made of cast steel, but instead cast aluminium to save weight -- a trick General Motors would copy decades later with the legendary LS Block, which are fucking obscene if you know what you're doing with them. No, I mean REALLY INSANE.
These monstrous 4x4s then got an evolutionary cousin:
The bouncer-climber. They got their name, because usually going up hill, if you run on slippery, wet or dissolving ground, you struggle and you're kinda fucked and the traction falls out of your ass and you're stuck. Well fuck.
The trick with the bouncer? You go up hard, and you come down hard. The resulting impact creates enormous ground-pressure, which sinks the wheel like a stiletto into a birthday cake (hence the thinner wheels) and that gives you mechanical traction and then you go forward, and you keep bouncing.
Unfortunately, this design has a limit:
All of its gubbins are down low where rocks can rip them to shreds, and they fight with the underside of the vehicle itself, because it can't elevate itself... They become breached due to the lack of containment.
Well, some insane nutjob decided to fix that.
He decided to pack the gear into an above wheel package, and have variable suspension.
Supposedly its climbing ability is mostly limited by the response-rate of its hydralics And the ability of the operators together, to use it properly. And that it should be able to go much faster through these environments.
Looking at real weird ground vehicles at absurd limits in sports is really insightful. That scenario screams for a suspension system that has a leg or hook, to use the ground normal itself as a means to pull itself.
Fair enough. So right now, we have something which is kinda sluggish, but can really go nuts with its wheels, and get over obscene terrain but it lacks a waistline or the automatic systems of its body to be counted as "an robot". Okay.
Well, an unexpected source decided to look into this problem, for their own purposes: Ground stability.
The issue with hydralics is they're fucking slow.
Strong, but so damn slow.
If someone can figure out how to implement a sort of bidirectional hydraulic rachet, which can hold positions in one direction, but accept play in another, and do so variable in both directions same as human tendons do, that would improve robotic motorics enormously and double as an incredible shock absorbing system.
I learned the hard way with the bone in my shoulder that our muscles are very weak, and our tendons act as counterforces, like the weights in an elevator
Its not that the motors in our robots are too weak and their bodies are too heavy (though it is partially this) -- its also that the successful work the motors are doing is constantly being fought, so the motors need ways to lock or hold positions -- usually with servos which use proportional control math.
That's the difference between goal and current value, how successful you're doing getting there, and how long until you get there in what's called a PID equation, which is then tuned by a computer, which is how our brains work our joints then through an inverse-kinematic solver which deals with endpoints or joint rotations as differing modes. We're seeing this here in this design.
I love LOVE that this thing has the missing part here:
A foot.
Yeah check that shit out.
Sure, it has to balance on its endpoints which kinda sucks (and here its balancing on a bucket too) but we're beginning to see the kind of variability and motorics needed, even if they're sluggish and slow.
What its missing is joint control and awareness, and obviously the agility to perform functional equlibrioception (balance percieved via pressure on your feet and joint positions) and proprioception (balance from the inner ear, and sight) and the live compensatory systems to do this.
How do we go upright, to walk?
Well, we do see this functionality in smaller robots, but we should focus on skating before walking.
See how it leans into its movements on all fours, and how it can upright? Do you now also see why I was questioning the use of a cab on top?
There are similar designs too, which can rotate the in-hub wheels to help control orientation when in the air falling, or logically in an upright design to help keep it balanced, and these designs are capable of lateral driving, too:
Its just two legs, with two very wide feet, but see how we're getting these kinds of agilities now? All because at this small scale, pure proportional control without mechanical fallback is 100% a viable control strategy, even though logically it doesn't really scale up efficiently, same as the mass itself doesn't scale up due to the square cube law which cubes the mass every time you square the size of an object.
But why would you do this?
Link to Part 2
I don't know how many people will find this, but I hope many many do! I'd love to talk with you all!
Raspberry Pi lends a hand to stop this robot falling over Kris Hauser is an associate professor of electrical and computer engineering, and of mechanical engineering and materials science at…
