The specific thing shown in this video is pretty useless, but there's likely to be a place for combat robots that can climb rocky mountainsides that armored vehicles would find impassable.
I'd point to tumbleweeds, and the Wheel Spider[1] as examples of wheel-like structures that evolved (in terrains where it makes some sense, of course - not so handy for a scree-littered hill, bog, etc.).
If it were a full featured "fiction" level Mecha, the main advantage is that it could crouch behind cover, go prone, jump obstacles and climb uneven terrain.
Add one more "tech level" to it and you can have a 1-on-1 mapping of the user's input and this allows you to have tanks that require almost no training to use allowing you to have your infantry be... well, big robots. The only advantage of that is if you can produce them faster than training would take.
The gamification is cool regardless of how ugly it’s executed. But if this goes anywhere it will most likely draw attention from less entertaining purposes.
The problem with actually running is that it likely will burn way more power at any speed where its trying to run, so this likely would be really boring to watch -- "he has 2 more seconds of his turbo to use... ohhh... and he's out.. gotta conserve power to avoid getting run down by the other robot at a blistering 4 miles per hour.."
With this level of control, you're probably right. But if you get the control solved and no human inside it would at least be as interesting as a horse or dog race.
The point of running and walking is simply that driving only works on sufficiently flat surfaces and is only efficient if the friction of that can be reduced, e.g. using flattening and asphalt. Running on sand or snow or otherwise loose underground, with big feet to distribute weight, is always going to be more efficient than rolling, we just can't control such systems.
But try doing a safari. A 6 hour drive across "flat" savannah, after which you look at the map and see you've travelled 30km, and used up something 1-2% of the lifespan of the 4x4 you were in (axle will need replacement after 50-70 such rides because it'll be shocked to pieces), that ride will cure you of the notion that driving is efficient.
And that's flat savannah, really, on photographs it looks really flat an driveable. And at the height of the dry season there are areas where that's even kind of true. Not many though. When you think about other biomes like forests, rocky mountainous areas, swamps (or really anything even a bit wetter than dry savannah, like almost all of Western Europe). Without roads, cars go nowhere.
But after dozens of billions of dollars of infrastructure investment per state, it can be tolerably efficient in densely populated areas. Even such efforts unlock something like 20% of the state at best.
Most people just happen to have spent their entire life in such an area and then extrapolate their direct experience to cover everything.
Then there's "reach". The truth is, even in dense cities, specifically built for car traffic, walkers can reach far more locations than cars ever will, even the smallest most agile car you've ever seen. And even there you see people using power equipment costing millions to compensate for the fact that cars, vans, ... can't navigate something like a building site.
I wanted to see it take multiple steps in a single shot. As an example, the videos released by Boston Dynamics are so interesting because you get a mostly unedited shot of the what the machine can do. This is just a disappointing commercial.
I'm highly skeptical that we'll ever see humans fighting against each other in giant metal deathtraps. We have demolition derbies but they're intentionally kept to a slower speed. Maybe some kind of sumo type fighting would be possible if it was highly restricted.
From this it barely walks, let alone runs. I thought maybe they were holding back for a shock finale race or something but then it switched to CG and that was it.
terrible. fire the marketing people. even if the target audience of this is kids (in the hopes that they will have purchasing power when this thing is ready for market) it is still terrible.
(don't get me wrong, i usually live for anything robotic, new tech, mech stuff but this is embarrassing).
excessively deep voice narrator - tick.
15 seconds of content stretched to a 2 minute video - tick.
All that effort, and it's a tripod-gait (one leg moves at a time) walk and a slow bound (front and back legs move as pairs) run.
It's mirroring the users motion, like the General Electric walking truck from 1965.[1]
There have been many fast-running robots. Boston Dynamics gets the most press, but there are others. I have an old patent on this from the early 1990s. It seems quite possible to do much better at legged running than people are doing, but there's no market and it's too expensive for the toy market. The Marines rejected the Legged Squad Support System, the militarized version of BigDog.
If you're going to build a big four-legged machine, it would be interesting to do it with double-ended air cylinders. Then you have adjustable compliance and energy recovery. A muscle can be thought of as a spring with an adjustable spring constant, zero point, and damping. Humans get back about 70% of running energy as elastic recovery. Cheetahs, about 90%. So energy recovery is crucial to fast running.
It's hard to implement that mechanically (there are mechanical kludges that do it, but not very well) but straightforward pneumatically. Carrying an engine and air compressor is usually too bulky for mobile robots, but in truck size, it would work. The control problem is different. You have to work in force space, not position space. It's not a good exoskeleton system, but it's a good robot.
When Google bought all those robotics companies, I was expecting great things. But nothing happened. Top management lost interest. Last year, Google sold off Boston Dynamics and Schaft to Softbank. Maybe they'll do something.
It's quite do-able today. When I was working on this in the early 1990s, gyros were big and expensive, as were accelerometers. Getting enough mobile compute power was hard. Nobody had 3D sensing good enough to profile terrain. Batteries were still NiCd. Motors were weaker. Power MOSFETs were bulkier and more troublesome. All those hardware problems have been solved. There's also been a lot of progress in control theory. Back then, it was mostly PID. Adaptive feedforward control was exotic and rare. Now advanced control is common. It takes smartphone levels of CPU power, which is a non-problem today.
32 comments
[ 4.4 ms ] story [ 63.1 ms ] threadA hummer weighs 2000lbs less and can go well over 80mph, what benefit does this concept have over armored cars?
Can go where no hummer has gone before ?
May also help firefighter carry heavy load for forest fire.
1: https://en.wikipedia.org/wiki/Wheel_spider
Because my inner child wants to pilot a giant robot, and that's that.
Add one more "tech level" to it and you can have a 1-on-1 mapping of the user's input and this allows you to have tanks that require almost no training to use allowing you to have your infantry be... well, big robots. The only advantage of that is if you can produce them faster than training would take.
In real life, right now, they simply looks cool.
The point of running and walking is simply that driving only works on sufficiently flat surfaces and is only efficient if the friction of that can be reduced, e.g. using flattening and asphalt. Running on sand or snow or otherwise loose underground, with big feet to distribute weight, is always going to be more efficient than rolling, we just can't control such systems.
But try doing a safari. A 6 hour drive across "flat" savannah, after which you look at the map and see you've travelled 30km, and used up something 1-2% of the lifespan of the 4x4 you were in (axle will need replacement after 50-70 such rides because it'll be shocked to pieces), that ride will cure you of the notion that driving is efficient.
And that's flat savannah, really, on photographs it looks really flat an driveable. And at the height of the dry season there are areas where that's even kind of true. Not many though. When you think about other biomes like forests, rocky mountainous areas, swamps (or really anything even a bit wetter than dry savannah, like almost all of Western Europe). Without roads, cars go nowhere.
But after dozens of billions of dollars of infrastructure investment per state, it can be tolerably efficient in densely populated areas. Even such efforts unlock something like 20% of the state at best.
Most people just happen to have spent their entire life in such an area and then extrapolate their direct experience to cover everything.
Then there's "reach". The truth is, even in dense cities, specifically built for car traffic, walkers can reach far more locations than cars ever will, even the smallest most agile car you've ever seen. And even there you see people using power equipment costing millions to compensate for the fact that cars, vans, ... can't navigate something like a building site.
Walkers are the future, at some point.
What's so special about it?
Cue guy putting on aviators
excessively deep voice narrator - tick.
15 seconds of content stretched to a 2 minute video - tick.
weak cgi - tick.
overly dramatic music - tick.
people striking heroic poses - tick.
and "XMech Racing League"? really?
There have been many fast-running robots. Boston Dynamics gets the most press, but there are others. I have an old patent on this from the early 1990s. It seems quite possible to do much better at legged running than people are doing, but there's no market and it's too expensive for the toy market. The Marines rejected the Legged Squad Support System, the militarized version of BigDog.
If you're going to build a big four-legged machine, it would be interesting to do it with double-ended air cylinders. Then you have adjustable compliance and energy recovery. A muscle can be thought of as a spring with an adjustable spring constant, zero point, and damping. Humans get back about 70% of running energy as elastic recovery. Cheetahs, about 90%. So energy recovery is crucial to fast running.
It's hard to implement that mechanically (there are mechanical kludges that do it, but not very well) but straightforward pneumatically. Carrying an engine and air compressor is usually too bulky for mobile robots, but in truck size, it would work. The control problem is different. You have to work in force space, not position space. It's not a good exoskeleton system, but it's a good robot.
When Google bought all those robotics companies, I was expecting great things. But nothing happened. Top management lost interest. Last year, Google sold off Boston Dynamics and Schaft to Softbank. Maybe they'll do something.
It's quite do-able today. When I was working on this in the early 1990s, gyros were big and expensive, as were accelerometers. Getting enough mobile compute power was hard. Nobody had 3D sensing good enough to profile terrain. Batteries were still NiCd. Motors were weaker. Power MOSFETs were bulkier and more troublesome. All those hardware problems have been solved. There's also been a lot of progress in control theory. Back then, it was mostly PID. Adaptive feedforward control was exotic and rare. Now advanced control is common. It takes smartphone levels of CPU power, which is a non-problem today.
But there's still no market.
[1] https://www.youtube.com/watch?v=coNO9FpDb6E