Boston Dynamics' new Atlas robot doing work.

[jonovid]

if the robot drops something? can it find the lost component on the floor and recover from the random deviation from its workflow, all without needing human intervention?
or will it system crash.

[SteveThackery]

if the robot drops something? can it find the lost component on the floor and recover from the random deviation from its workflow, all without needing human intervention?
or will it system crash.

We obviously don't know, because we're all looking at the same video. What is your point?

[xrunner]

if the robot drops something? can it find the lost component on the floor and recover from the random deviation from its workflow, all without needing human intervention?
or will it system crash.

I don't know, but I can safely say this. This is the worst that these robots will be able to perform. It will only improve.

[jonovid]

if the robot drops something? can it find the lost component on the floor and recover from the random deviation from its workflow, all without needing human intervention?
or will it system crash.

We obviously don't know, because we're all looking at the same video. What is your point?

my point I would to see this complex self balancing humanoid robot do more then just follow a basic set sequence like a car manufacturing robot , jukebox disc loading mechanism or a cam following 1800s automaton.  IMO complex multitasking is the essence of a humanoid robot.

[lokin4areason]

can see the military and the police having a go with this along with the manufacturing/ construction sector in eliminating manual labor aspect

[tom66]

Even if 10 robots need one handler to sort out the odd issue, it's a major reduction in workforce requirements, with all the associated social upheaval.

I suspect a 100% automated factory is decades off. There are some limited examples of this, but only very specific work that lends itself to automation.  General automation by autonomous robots is a very hard problem to solve.

[Bud]

can see the military and the police having a go with this along with the manufacturing/ construction sector in eliminating manual labor aspect

A walking robot is a niche use case. It is much more effective if it rolls, especially in controlled environments like factory floors.

[SteveThackery]

my point I would to see this complex self balancing humanoid robot do more then just follow a basic set sequence like a car manufacturing robot , jukebox disc loading mechanism or a cam following 1800s automaton.  IMO complex multitasking is the essence of a humanoid robot.

Why do you require "complex multitasking" from a humanoid robot in particular, rather than from all robots?  Surely there will be applications which require complex multitasking, and others which don't. The shape of the robot (box on wheels, or humanoid) shouldn't affect how much multitasking is required.

I'm just saying that those questions you ask are totally legitimate, but are questions that apply to the whole field of robotics, not specifically to Atlas.

[BrianHG]

Atlas doing exercise/pushups.  Also, his face plate is off revealing a bit underneath.

https://www.youtube.com/shorts/asHc4a55JX4

[SteveThackery]

Atlas's walk still isn't quite right. It seems never to lock out its knees, which most humans do briefly when walking at a moderate pace. Its permanently bent knee walk undermines its overall - astonishingly good - performance.

[eutectique]

As a side note, this is something the "realistic humanoid robot" on this forum will never be able to do. 

Just give it some time.

[pcprogrammer]

As a side note, this is something the "realistic humanoid robot" on this forum will never be able to do. 

Just give it some time.

Doubt it, but he does have the lord on his side. 

Atlas's walk still isn't quite right. It seems never to lock out its knees, which most humans do briefly when walking at a moderate pace. Its permanently bent knee walk undermines its overall - astonishingly good - performance.

It is also loud. Must be the flat feet Atlas has. But certainly impressive how it goes out and in from to standing position to do the push up work out. They should try weight lifting, also a balancing act.

[Siwastaja]

There is something to Boston Dynamics which makes them appear more credible than nearly anyone else:

1) They have chosen to build robots with really challenging mechanical... dynamics. The balancing actions they show cannot be faked with remote controls. They have to be truly autonomous features.
2) They are not just limited to stability of simple movements. Instead, they use their... dynamics very creatively in complex patterns. Again, can't be faked with remote controls (well, maybe except to give very high-level guidance, "do trick X now").
3) They demonstrated all of the above more than 5 years ago. From the current demonstrations the dynamics have not improved much, which is clearly because they were so impressive years ago already there is not much to improve. That very likely means they have been using their same engineering skills for something new, at a different level.

Besides, the demonstration they show in the video posted is not that impressive. I mean, if it was faked, they would have aimed higher. It is just right level of impression to be plausible.

There is a lot of "fake it until you make it" in robotics, and as such a lot of distorted expectations what can be and cannot be done. Having worked in a failed (seriously under-resourced and ran by... let's say unconventional opportunists) robotics startup some 5 years ago, I saw this pretty closely. Now when I see "autonomous projects" like delivery robots I'm nearly 100% sure they are at least assisted if not completely remote-controlled from some Indian call-center type cheap labor operation. It's not surprising Elon Musk does the same.

But I do believe that Boston Dynamics is one of the few who really fakes order of magnitude less than industry median. Possibly even near-zero faking.

[Njk]

Sorry for a low-tech question, but how it supposed to be powered?

Early the same company demoed a robotic donkey that can carry various payload, e.g. ammunition for a squad of soldiers. What was funny is that the donkey was noisy like a motorcycle because it used an ICE for electricity generation.

This device is not created for heavy lifting, but it's very agile so it also requires lot of energy. No problem to keep it powered for five minutes of shooting, but what's next? Even a living thing can't work forever, it need a considerable time to rest and to recover their energy by digesting the food. No way to beat them on efficiency, they're created by a billion years of evolution.

Any man-made machine is much less energy efficient. So if one is going to build an army of that robots, that army will require much more energy than a human army. Where to source it? Simply because of that it's obvious a devastating effect from the robot's population will be much more serious than that from a human population of equal amount. Although it's interesting business for someone.

[SiliconWizard]

Well, interesting question. I'd be really curious to know what the average and peak power consumption of these robots is.

[CatalinaWOW]

Energy consumption of robots isn't so clear cut as that.  I'll stipulate that joules/motion might be better for humans, but humans consume a lot of energy just heating the body to operating temp.  Robots do have to do some temperature control, but have a huge advantage in this part of the power budget.

[Siwastaja]

Any man-made machine is much less energy efficient.

I don't know, these could be surprisingly energy efficient. Like, the motors turning the joints regenerating to stop. Muscles can't do that, they dissipate into heat when "braking".

Amount of physical work done in the opening video is near-zero; it's not working at speeds where aerodynamics are a big issue. Assuming the joints have good bearings (and why wouldn't they), and any electric motors and gears are in modern >80% efficiency, I see no reason a 2-3kWh li-ion pack would not enable it to work for a long shift.

[Njk]

Energy consumption of robots isn't so clear cut as that.  I'll stipulate that joules/motion might be better for humans, but humans consume a lot of energy just heating the body to operating temp.  Robots do have to do some temperature control, but have a huge advantage in this part of the power budget.

The whole point of having a robot is to let it to do more work, increasing the duty cycle to maximize ROI. So the robot will have a cooling problem, not a heating problem.
But what's more important, from the environmental perspective, such autonomous robots seem like a plague worse than cars (unless the mass production is banned, which is unlikely). A way to hell

[Siwastaja]

The whole point of having a robot is to let it to do more work, increasing the duty cycle to maximize ROI

Maximization of duty cycle cannot be the whole point of having a robot. Maximization of ROI is, fair enough, but is not the same as maximization of duty cycle at all. For example, if you can reduce the duty cycle by 30% and halve the cost of the robot, then it makes more sense to do so, and have two robots instead.

Duty cycle is just a linear coefficient for availability. Four robots working at 25% duty cycle each could be well acceptable for most purposes, all that matters is cost effectiveness of the whole (how much work they do for the total cost).

From battery charging and discharging viewpoint, I would expect reaching 50% duty cycle is trivial and even 75% not difficult. Compare that to roughly 25% of a human being (6 hours of actual productive work per 24hrs).

[bdunham7]

From battery charging and discharging viewpoint, I would expect reaching 50% duty cycle is trivial and even 75% not difficult. Compare that to roughly 25% of a human being (6 hours of actual productive work per 24hrs).

Surely just having swappable battery packs is a simple solution to get near 100% duty cycle.  There are electric power tools now that are replacing gas-engine models even in high-demand applications like concrete saws, chainsaws, etc.  For example, the Milwaukee MX Fuel 12.0AH (72V) has about 1.3kWh of energy and can charge in about an hour.  With one or two spares cooking on a charger and a robot that can change its own battery, there's no reason it can't work all day even at a pretty high pace.

[Njk]

Duty cycle is just a linear coefficient for availability. Four robots working at 25% duty cycle each could be well acceptable for most purposes, all that matters is cost effectiveness of the whole (how much work they do for the total cost).

Sure, and that will make the robot sellers happy. For all the others, it seems less than ideal because it not only quadruples the required garage space. It's not a computer game, where you can replicate an object by the mouse click. Production of every robot implies the use of manufacturing space, energy, raw materials, etc., and the required efforts may not be always measured monetarily (it's not a hard unit anyway). For instance, it can be suddenly found that the most suitable materials belong to the rare-earth category. They're hard to find. Many people will object against the plans to undermine their houses for that materials, and for a good reason.

From battery charging and discharging viewpoint, I would expect reaching 50% duty cycle is trivial and even 75% not difficult. Compare that to roughly 25% of a human being (6 hours of actual productive work per 24hrs).

The energy density makes huge difference in possible applications. Different minimal density is required for different machine types. A wind-powered floating machines are known from the ancient times. A battery-powered rolling machines are in big fashion now. A battery-powered toy flying machines are known, but a battery-powered commercial airliner does not exist. It remains to be seen if it's feasible. It seems the walking machines are sitting somewhere in-between. Everyone knows about the batteries, but remember, the robotic donkey was powered by ICE.

[CatalinaWOW]

My point about the human consumption of energy was not pointing to a duty cycle issue in robots.  I see no obvious reason why they would be thermally limited to less than a 100% duty cycle.  Radiators aren't that hard to add if required. 

Sorry about the use of non metric units, but here in the US calories are THE unit of choice for evaluating human energy consumption.  Basal metabolism (just operating the body with no physical activity (keeping it warm, pumping blood and air and thinking) consumes something on the order of 2000-2300 calories per day for the average size male.  Heavy physical activity can roughly double that, indicating that at best the human body is 50% efficient in generating physical work.  The actual number is surely lower because there is almost no recovery of energy except for the spring action in the legs during walking and running and there is friction everywhere.  Cold weather makes it worse, with up to an additional 2000 calories a day to stay warm in extreme conditions.

[jonovid]

must have slip rings in its waist and hip sub-assembly. impressive engineering 

[SiliconWizard]

But can it play Doom?

[BrianHG]

But can it play Doom?

You bet the military would like that!