what's the biggest obstacle right now for making bipedal robots that can do complex tasks?
Jews
Proper balancing and quick movement.
Basically we need synthetic muscles.
>>51232309
are electroactive polymers viable?
batteries
I dont know why you'd even bother trying to make a bipedal robot unless just for familiarity. There's no practical reason for it and it just seems like more work than its worth
>>51232309
We need tiny muscles. Humans have shitload of different muscles that all work together to balance you.
But I think it's the wrong direction we're taking bipedals. We should optimize their robotic nature instead of trying to emulate humans.
>>51232488
>>51232309
Nasa's robots have muscles that are perfectly applicable in place of humans. The only actual problem is software.
>>51232525
>perfectly applicable in place of humans
You fucking wish kek
>>51232488
but we already did optimize their robotic nature for factories in the 60s. why go back to that?
>>51232245
Tiny human brains
>>51233079
Tiny
White
Keks
>>51232525
The balancing/software aspect is way ahead of the physical limitations we're facing.
NASA's robots have 1 "muscle" in places where our body has 20, and these actuators are either too bulky or too slow. We need the equivalent of a muscle, compact and instantly adjustable to a precise level of contraction.
>>51233526
OP here. like >>51232321? afaik they pretty much fit the bill outside of elastomers capable of handling a lot of weight. the beauty with them is that the actuator is a coat of carbon on a polymer that you intend to flex.
>>51232450
Our entire world of tools and machines is essentially built for use by humanoids. Are you actually this dense?
Processing power and a good AI. Right now robots are being forced to do processing work on nothing more than some shitty intel atom or arm cpu. Simple things like walking require being able to balance properly, predict where the next step should be without looking at the ground or feet, and be able to make split second changes to avoid falling on the ground like when you step on a rock or walk on an uneven surface. Shit is very computationally intensive.
Even with powerful processors doing work off site and relaying it back to the robot our AIs are far too primitive and glitchy to be useful at the moment.
All this will change soon however as demand for humanoid robots to replace laborious jobs grows and robots become cheaper to manufacture. Right now pipedal robots are nothing more than fancy and expensive trinkets.
>>51232245
WHY THE FUCK DO THEY HAVE TO BE BIPEDAL? THE FUCK YOU NEED LEGS FOR ANYWAY?
Most of the things we want robots to do, can be done with a wheeled platform with arms. But we can't do these task because we haven't solved the problem of grasping and manipulation. So even if we solved "bipedal locomotion" today, our robots wouldn't be able to do shit. Grasping and manipulation is much harder than walking to solve. Walking is mostly an engineerin
>>51232309
Nope. The problem is the torque density of the electrical actuators we have. We have actuators with amazing power density and efficiency, but low torque density.
>>51232488
>>synthetic muscles
why the fuck is everyone obsessed with synthetic muscles? We don't need our robots to have a ton of fucking muscles because unlike nature we can make rotating parts and get rotational motion out directly.
fuckers, I work on artificial muscles and they certainly are not gonna be the panacea that fixes all problems in robotics. We are still gonna need rotary electric actuators for robots.
>>51232321
>>are electroactive polymers viable?
Hahaha no. Polymers are spaghetti at the nanoscale. Spaghetti is terrible for making stuff that moves. If we really want to make artificial muscles, we're gonna need to make highly organized nanoscale systems.
>>51232386
Not a problem if we actually worked on making efficient actuators.
>>51232525
>>Nasa's robots have muscles
HAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHHAHAHAHA. No. Not a single one has muscles. Robonauts 1 and 2 use all electric motors, RoboSimian uses electric motors, Valkyrie uses electric motors.
Valkyrie is an interesting case though, you know the problem with torque density? Well Valkyrie uses some new actuators that have a high torque and power density. But these actuators are really springy. They need to be springy, because with springyness you can control the torque they apply with a higher bandwidth. But you need special control approaches to deal with the springyness
>>51234103
what about HTS motors then?
>>51234103
>WHY THE FUCK DO THEY HAVE TO BE BIPEDAL? THE FUCK YOU NEED LEGS FOR ANYWAY?
>
>Most of the things we want robots to do, can be done with a wheeled platform with arms. But we can't do these task because we haven't solved the problem of grasping and manipulation. So even if we solved "bipedal locomotion" today, our robots wouldn't be able to do shit. Grasping and manipulation is much harder than walking to solve. Walking is mostly an engineerin
Isn't bipedal movement more efficient for close quarters space? Like buildings and whatnot?
>>51233811
And by law, much of it is required to be wheelchair accessible.
>>51234046
>>Right now robots are being forced to do processing work on nothing more than some shitty intel atom or arm cpu.
U serious mate? Pic related uses Java(muh agile development) that has to run on a single processor core because this causes less jitter.
>>balance properly, predict where the next step should be without looking at the ground or feet,
aren't that complicated really. The problem is while we can figure out the best thing to do in a split second, our actuators can't react fast enough. At this point it's more of a hardware problem than a software problem.(that is if you are trying to do this without looking at the ground.)
But yeah we're gonna need AI to solve the ridiculously hard problems of grasping and manipulation. Reaching into a clutter cupboard to grab a cup of ramen noodle is REALLY FUCKING HARD. Because now instead of just planning how the arm and hand is going to move, you have to plan about how all the shit in the clutter cupboard has to move. We're gonna need better AI approaches than just randomly generating and testing robot configurations to solve this(it's called RRT).
Goddammit.
>>51232245
You want to know when we'll get robot waifus, admit it.
>>51234636
not gonna lie, it's a portion. probably due to nofap november homie. it's making me debate on switching majors to accelerate the progress so i can get my own android 18, even if its not super complex AI just the body with some presets.
>mfw
>>51235019
Switched to ai last month. No regrets :^)
>>51232245
AI isn't sophisticated enough for robots to do human level tasks yet.
>>51234288
>HTS motors
sure if we can solve the problem of making a room temperature superconductor. And of course it would really help if said room temperature could be formed into really thin wires so we could get a lot of turns... Of course a magical infinitely torque dense and power dense actuator(available in simulation!) won't solve all your problems. You also need fine force/torque control and if you really want to be efficient you need your legs to act something like pogo sticks.
>>51234492
>>Isn't bipedal movement more efficient for close quarters space? Like buildings and whatnot?
not really, because buildings are required to be wheelchair accessible. And if you really want to do stairs or weird rough terrain stuff check out NimbRo rescue's mamaro robot. It's a wheeled robot that can drive a car and climb stairs.
Fun facts about pic related: Pic related is PR-2. It has been able to operate for 13 days 2 hours in an office environment completely autonomously without any human help, even plugging itself into to recharge. Over this time period it traveled over 138.9 km. It would have been able to operate for longer if ubuntu hadn't encountered a kernel panic error.
>>51235213
>not really, because buildings are required to be wheelchair accessible
Someone's never left the US.
Here's my thoughts on walking
The knee movement should be easy to replicate, because it flexes in two directions. The hip should be fairly easy too, because it only needs to flex fpur ways. The ankle and toes are the super complex part we need to figure out before shit starts to walk
>>51235251
3rd world shitholes are not rich enough to afford robots, so this isn't a problem
>>51232245
There is no point on making humanoid robot
>>51235128
yeah but i'm looking for the physical side, i was thinking material science but i might be off. something to get my handle on servos and elastomers if you catch my drift.
>>51235393
a humanoid robot would allow otherwise obsolete tools and tech to be used with greater performance, reducing landfills.
>>51232321
>electroactive polymers
A question worthy of an answer.
The two research problems are they use way too much power and they fall apart easily. Especially when you push 2kw through them to make them move.
>>51235472
Just made this webm, but hit post by accident.
In summary its simpler just to genetically engineer our own slave race
>>51235510
Why won't we just use democracy to enslave 49% of the population?
[spoiler]Is it because 1% used corporatism to enslave the other 99%?[/spoiler]
>>51235472
>>51235494
that is awesome. what do i need to take in undergrad to pursue studies in this to improve the technology? it is so up my alley. i would definitely wanna try solving those problems if it means we can make limbs without bulky systems.
>>51235556
Materials science for sure.
https://en.wikipedia.org/wiki/Dielectric_elastomers
These things are basically capacitors, but are gonna need next gen materials to make them work. All the nano stuff you hear about.
>>51232245
no use and too expensive
>>51235621
>no use
>>51232488
This. What is the point in a biped when we can make perfectly functioning quadrupeds that can do the same task? It's liek if the inventors of the car said, fuck wheels we gotta make it move like a horse.
>>51234568
This guys has it right. If we had smart software we could compensate for crappy actuators. Unfortunately robots are retarded. Like on the level of a retarded 2-year old retarded.
Robots. Have a hard time perceiving individual obects in a cluttered unstructured environment. Now try solving that problem in realtime with a computer you can fit inside a robot. Even if you used offboard computing and somehow managed the latency issues to solve the perception problem with CNNs you still have to solve the problem of understanding high-level human commands and planning and execution in a cluttered environment. Really the physical robot isn't so much the problem as the fact that we dont have good enough algorithms for planning and perception. I would argue that humanoid robots are expensive mostly because they are research toys not because we couldn't build a better one. We really dont have the algorithms to make them useful outside of the lab.
>>51235677
>Billions of dollars and decades of R&D just because some fat ugly autistic fuck can't get a girlfriend
>>51235724
that isn't the same at all. we have human tools all over the place, horses aren't. why even say this? troll/10.
>>51235424
and most tools we use with our hands rather than our feet, which means a wheeled base with a pair of arms would be just as useful
>>51235510
kek no. Genetic engineering basically isn't. Right now "genetic engineering" consists of put some gene into something and see if it works. But most of the time it does not. We can't predict what's going to happen before hand. And we aren't very good at this insert genes into stuff part either.
Trial and error is all good and fine if you are working with bacteria because they grow fast, but impractical for primates which have a long gestation period. Imagine trying to debug a program that takes 9 months just to compile!
Second if you really want to make a good slave race that's not gonna rebel, you need them to be stupid. Which gives you about the capabilities of the mentally disabled. So even if we did manage to genetically engineer a slave race, it'd be awkward and weird to have them around.
>>51235748
Don't forget replacing soldiers.
>>51235800
BETA UPRISING
>against the law to clone people for sexual purposes
>perfectly okay to build people robots for sexual purposes
I'm sick of being brought down by the man.
you all should check this out if you haven't already.
https://www.youtube.com/user/BostonDynamics
No incentive to do so.
Chinese kids are cheaper for the foreseeable future.
Neuroanatomical models we need
>>51235800
why soldiers? just nuke 'em all
>>51232450
>>51234103
>>51235393
>>51235724
How do you make a robot sexy and lovable when it ISN'T bipedal?
>>51235909
>just nuke 'em all
Because oil
>>51232245
Both not understanding how to implement the complex task (the SOTA on making coffee currently takes over half an hour and spills ingredients), and actual motion.
>>51235800
What is this from?
Asking as someone who has seen Terminator 1 and 2 dozens of times
>>51232245
refugees welfare drain and the subsequent impact on fund availability for r&d in areas of science
>>51235615
gonna study material science. wish me luck guys!
>>51235922
This anon isn't wrong entirely. It also just makes sense since we could produce one robot that can do multiple tasks and make the programming take care of how to do all those tasks rather than manufacturing a different robot for every task. Once you have software it costs you pretty much nothing to put it on every robot.
Unfortunately it turns out that being able to solve that kind of high level complex problems is much more diffcult than manufacturing a roomba style robot for everything you don't want to do yourself.
>>51236015
on all fours. tires on elbows and knees.
>>51235973
T2 teaser trailer. Most people saw it probably on the Total Recall vhs.
>>51236040
great now i'm picturing roomba lolis on all fours dashing around. my sides.
>>51236010
Pick a college with good government contracts like CMU and you will be golden.
>>51236124
i'm in canada actually. it seems that my nearby universities don't offer material science, there's one that offers nanoscience but i think that's a different specialty. not really sure what to do...
>>51235745
>>If we had smart software we could compensate for crappy actuators
Compensate? Did I say compensate? What I meant to say was that better algorithms could make our bipedal 2 armed robots about as capable as an old lady. Old ladies may be smart, but they still fall. We still need to solve the problem of actuators.
>>perception problem
>>have to solve the problem of understanding high-level human commands and planning and execution in a cluttered environment
Unfortunately, a vast majority of these things can be done offboard with latency. You don't need to know that the object in your way is a chair every second because SLAM works pretty well. You don't have to understand what a human is saying in less than a second(look at fucking Siri). It doesn't matter if a robot stares at a cupboard for a minute before it receives a plan on how to get ramen.
Latency isn't as much of a problem for high level planning. This is a problem because it means a company like google could sell this high level planning as a service. And they might be able to make it work really really well, because they'll get huge amounts of training data and cases. Instead of planning the optimal grasp for a coke can, they could just reuse optimal grasp they generated before and maintain a huge database of said grasps.
This is literally the most botnet thing ever. You'd have a robot with advanced sensors and actuators sending environment data off to google for object recognition and planning. The object recognition part alone is pretty scary.
>>51235922
>>51236258
Uwaterloo, utoronto, udem, mcguill?
>>51236258
kid. I work on artificial muscles. Electroactive polymers are not the way to go. The future of artificial muscles is in rotaxanes.
>>51236310
i'm in ottawa, closest would be Carleton and uOttawa, then either Waterloo or Mcgill. Carleton has the nanoscience and i'm enrolled in CS at uOttawa but not taking any sessions.
>>51235922
I find writhing masses of fractally branching manipulators to be sexy.
http://www.frc.ri.cmu.edu/users/hpm/project.archive/robot.papers/1999/NASA.report.99/9901.NASA.html
>>51236359
pssst... the future of artificial muscles is rotaxanes. One of the lead researchers doing this work is in Canada.
>>51236348
>>51236348
isn't that super small scale though? i thought that was just for molecular machines.
>>51236471
Canada is pretty much at the forefront of robotics and CS these days for some reason (AI, quantum computing and cryptography especially).
>>51236619
>for some reason
probably because it's too cold most of the year to be attractive to the sjw hippies
>>51236474
not if you make chains of them.
>>molecular machines.
that's what they are.
>>51236619
Sure thing bub, I mean just look how well the Canadian robot team did in the DARPA robotics challenge. Pic related it's the Canadian robot team.
>>51236826
there's no pic...oh you dastardly fellow you.
>>51236741
But there are tons of SJW in canada.
>>51236826
>participating in challenges and revealing the direction of your research and level of progress
that's not how you get ahead in business or science bub
>>51236900
did not know that, I thought america light was free of them
>An actual technology thread on /g/
As I live and breathe! I forgot what they were like. They used to be decorated with loli but this is fine too.
>>51236262
My point with the compensation part is that we dont have good enough motors because we really have nothing to do with them. If we had the software we could invest heavily and iterate on our current motors to get them up to speed while still having a viable product. Maybe our robots would be easy to trip but they might stills be useful for something. A "smart" granny can be useful in getting the money flowing a brain-dead athlete probably isn't.
On the algorithms part, it's not just the latency. Planning is a hard problem. If you have a room filled with shit like most people it's even harder. Not just that it's a n-dimensional problem where n is your number of joints. It's a hard problem as in even all the googles and AWS might not be enough to understand->plan-> actuate for a single robot in a cluttered environment. Sure Siri can listen to you and google and set your schedule but telling a robot to do the dishes is orders of magnitude more computationally expensive. At least with current methods. Even your RRTs will get stumped when you have so many objects adding dimensionality to your problem.
>>51232525
in space maybe. but under standard Earth surface gravity, they wouldn't work even as well as industrial robotics--which are specifically designed to do useful work in that condition.
>>51235922
It's possible, fuck, I find tones of stuff attractive, but there's something about a self conscious biped that never thinks she's good enough for you that turns me on to no end.
Pic semi related.
>>51234103
whats keeping us from using hydraulics like the terminator?
>>51232525
Software and sensors.
I know that what works for biology doesn't exactly translate to machines, but I guess both pressure and tension sensors (along with standard gyroscopes) might help, since we work the same way.
It's hard, I know. You need a team with good knowledge in mechatronics, programming, and average notion on the peripheral nevrous system and muscle physiology.
The biggest obstacle is the lack of genetically engineered men small enough to fit inside the robot's heads anon.
>>51240075
>>>/containment/
>>51235922
>sexy
probably not.
>lovable
sure, why not?
>>51232245
no monopoles
>>51234492
There are individual bots with obscene "hand eye" coordination, dexterity of digits, etc.
I've one that always catches a tennis ball no matter how you throw it.
Another that can tie knots with two fingers and juggle chopsticks like a cokefiend giving a handjob, no reason this couldn't be combined.
>>51235406
mech. engineering w/ mechatronics?
>>51232245
Shit's complicated.
>>51232245
Humans
>>51232245
The KFC Double Down Sandwich.
Modern robotics can't handle that shit.
Just fucking look at it.
Two crispy chicken fillets.
Bacon.
Two kinds of melted cheese.
You got that? Two fucking kinds of cheese.
The T800 don't know what the fuck to do with that shit.
>>51244179
Howso?
>>51232245
Political Motivation & funding. If Murrica had as much passion to produce such a thing as the Nips do, it'd be a done deal already.
>>51235251
If companies can afford robots they can afford to tailor the robot work environment to be wheelchair accessible
>>51244488
that is disgusting
Women
feminists are lobbying behind the scene to defund the research institutions and scientists working on robots because as soon as they perfect bipedal robots, someone will start making sex robots, making all women redundant and unnecessary
>>51246593
Won't VR be better for this then robots and probably come sooner?
>>51246569
You're a fucking moron. The rest of the world is not nearly as politically correct or catering as America is.
>>51246314
I read an article somewhere that compared American technology industries with Japanese industries. It basically hypothesized that American companies were more focused on how the could monetize their products, while Japanese companies were focused on being "the best" at whatever project they worked on. Kind of like a pride issue. The nips compete with each other not for profit, but for bragging rights or some shit.
The ending theory was that most American companies lacked the sufficient inspiration to keep moving forward because they can so easily milk products & services for all their customers' money. Japanese companies on the other hand, saw no problem going into massive debt just to build their magnum opus and rule their field for a number of years.
07478354304
>>51246593
Not even behind the scenes. Pic related.
>>51246619
Whatever physical component is attached to the VR will be labelled a robot. Women don't discriminate when their monopoly on men is being threatened.
>>51235922
>Not finding Blade Wolf sexy as fuck
>>51246649
>Japanese pride
Oh I have no doubts what you're saying is true. Companies there literally form 100-year business plans while the greedy cock-monglers here panic if a single quarter shows a little downturn.
Nevertheless Murrica has way more money and can mobilize the shit out of things in short order despite the lumbering government here. Japan basically moves like a slug by comparison.
For something this complex and controversial however, it would probably take a major war to ever make it a national priority (ala Apollo), and even then I don't think humanoid robots as luxury items would ever qualify. There might be a rapid acceleration of military ones however, and that would have a trickle-down effect after things cool off (assuming things COULD 'cool off' heh).
>>51246684
What's totally asinine is I don't see her and her ilk lobbying for the 'rights of automobiles, or refrigerators, or industrial robots'. Surely us CIS scum are bending them to our every little whim.
I ask where's the justice when you can force you're filthy lewd dirty clothes into the Maytag, and it gets no say in the matter?
>>51241781
>>51235406
>mfw i'm doing mechatronics with a mechanical engineering minor
I'll build my own robot in 20 years. You fucking watch me.
>>51249262
>20 years
better get started today anon. gud luck.
Upright bipedal weapon? Too much mgs
This one would make a good start?
>>51242107
I'd love to see what the fuck was going through that things head at the time.
So many of the bots in that video seemed physically competent, just bumfuck retarded mentally.
Like not checking if you're actually holding the doorknob or not when you use the doorknob as your support.
Also all those fucking Japanese ones with them ears.
>>51252160
Ones and zeros mate, 1's and 0's.
>>51252160
looks like it tried to take a step forward with both legs at once..
>>51252260
kek
>>51253049
>take a step forward
>both legs at once
Fuck me, it makes sense but it's unfathomably hilarious.
Someone forgot a fucking pause in there.
>>51232245
>NSA acts like it has difficulty getting files from apple
>all crooks switch to iOS
>NSA wins
classic honeypot
