Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Open Problems in Robotics

scottlocklin.wordpress.com

Post reply on HN
Print view View original

211–220 of 232 posts

Re: Open Problems in Robotics

#211

Post by bluGill »

sfvisser wrote: post #16

Autonomous robots sound cool, but there are so many problems that could benefit from robotic automation that don’t require full autonomy. In controlled environments where we know the entire state of the (local) world including the position of our robotic swarm we can deterministically plan a motion and leave the world in an expected end state. Maybe use a bit CV to make sure our assumptions are right, if not hit the…

As an employee of John deere I can tell you that we are currently shipping most of the automation you suggested. Humans are still in the tractor but they often are not driving them.

Re: Open Problems in Robotics

#212

Post by blauditore »

> 6. Depth estimation

At least indoors, this was already handled pretty well 10 years ago by Kinect, and there are many somewhat robust approaches based on binocular vision. Not sure if this qualifies as "very much an open problem".

> 9. Scene understanding

The mentioned example is not the best one - anticipation of collisions has been a feature in cars for many years now (usually some warning system, sometimes with automatic emergency breaking). But it's a very constrained problem; more generic scene understanding is of course still very difficult.

Re: Open Problems in Robotics

#213

Post by snovv_crash »

cpgxiii wrote: post #26
krisoft wrote: post #15

Ugh. This is making a hash of it. What is true: The understanding of laymen about what robots can do is very much out of touch with reality. Making real world robots is very hard. I think it's movies which made people think that any 6 year old can just plop together a C-3PO. What is the hash then? (I omit the points where I lack experience.) 1; Motion planning "Even things like a model of where the robot is, with res…

Depth estimation I wish it were as solved as vendors would like to believe it is. If you want a sensor for medium-range applications, say order 0.5m to 10m depth with better than 1cm accuracy, that works indoors and outdoors, with reflective and untextured objects, doesn't interfere with other sensors of its kind, it simply does not exist. Traditional active and passive stereo is OK provided you have texture, ToF so…

I'm in the same boat. I'm now leaning toward the idea that single-shot depth estimation isn't the way to go. Throwing away all of the rich motion and flow information from previous frames seems like such a waste when we could be using it to update a dense model instead.

Re: Open Problems in Robotics

#214

Post by cpgxiii »

sudosysgen wrote: post #142
HALtheWise wrote: post #138

Earlier quoted context omitted.

The parent comment was clearly written by someone with no experience in robotics. In fact, every single problem on this list is "solved" in the sense that there are examples of systems that perform it in some limited context. Generalizing those subsystems to handle the difficult edge cases and integrating those subsystems together is 99% of what makes robotics as a whole difficult, and it would be similarly deceptive…

I do have experience in robotics. There exists robust solutions to depth estimation and position estimation that are deployed in the real world, in difficult applications where they have been generalized to work with any object you can draw a box around. There is a difference between a problem that is solved in theory and a problem that is solved enough so that you can buy polished products that implement a solution…

Structure from focus is an established technique that can produce quite good results, but it's not quite as easy as the existence of good autofocus in cameras would suggest. SfF techniques often operate the the inverse of an autofocus system - rather than focus on a specific point to figure out depth at that point, you run through a full focus cycle and then for each area of the image, you find the focus setting that made that area in focus.

The challenge there is that running through a full focus cycle and capturing frames as you go is quite time-consuming and introduces serious artifacts if anything in view is moving. Phase-detection autofocus is so fast in part because it's only using a sparse set of autofocus points, rather than the full sensor resolution.

The other major problems with SfF are the same the photographers encounter all the time - the depth value is only as accurate as the depth of field will allow, and shallow depth of field requires wide apertures (and is generally easier on larger sensors, as well). Large lenses and sensors may work well for stationary scanning applications, but struggle to be useful on robots.

Re: Open Problems in Robotics

#215

Post by malcolmwhite »

MichaelRazum wrote: post #172
mchusma wrote: post #126

I work in robotics. The problem for pretty much everything is one of scale and commercialization. Every one of these issues has been addressed, but almost none have a good and bundled solution. It feels like computers in the 70s. All the pieces are there but there aren't mass produced PCs yet somewhat because there aren't the suppliers to make things easier. This is changing fast. For example, SLAM has many things th…

What do you think about RL learning approaches? Just wondering if that stuff may work in practice. If I'm not wrong Sergey Levine said, that the only problem is sample efficiency. So you would need to simulate the real world, or let your robot brake 1 million time dishes till it learns.

Not OP, but where I work (Fox Robotics) we leave RL to demos and researchers. It is very promising, but less understood and less proven, so we aren't in a hurry to design any products around it.

Re: Open Problems in Robotics

#216

Post by cpgxiii »

iamgopal wrote: post #50
modeless wrote: post #13

The objections to neural nets as the solution to all the problems on his list are the same as the old objections to neural nets as the solution to computer vision, speech recognition, translation, playing Go, etc. The objections will fall in the face of overwhelming evidence that neural nets simply work better than other approaches to these types of problems. For a long time software was the reason robots didn't work…

Any research/potential example of such actuators ?

I would say some of the most promising work right now is in quasi-direct-drive actuators, basically low-gear-ratio brushless motors. Standalone electric motors are basically linear in terms of electrical power->torque, and even with a gearbox, provided the gear ratio is low enough, you can assume near-linear behavior. Brushless motors tend to be easier to design around specific size and shape constraints, and often run slower than brushed.

The "secret" to a lot of this actuator work is that outside of highly repetitive industrial tasks, actuators are rarely running at full power so you don't need to design them (and their cooling) for continuous max power operation. Somewhat like human muscles, you can pick smaller motors and/or lower gear ratios that cover the normal cases, and still be OK with rare overloads so long as the overload isn't long enough to cause overheating.

Re: Open Problems in Robotics

#217

Post by mchusma »

MereInterest wrote: post #130
mchusma wrote: post #126

I work in robotics. The problem for pretty much everything is one of scale and commercialization. Every one of these issues has been addressed, but almost none have a good and bundled solution. It feels like computers in the 70s. All the pieces are there but there aren't mass produced PCs yet somewhat because there aren't the suppliers to make things easier. This is changing fast. For example, SLAM has many things th…

> Teleportation was "possible but hard", but Freedom Robotics has a good solution now that mostly "just works" For somebody not in robotics, what does "teleportation" mean in this context. I assume that it doesn't mean the Star Trek style beaming that comes up from a Google search.

*yes, teleoperation (autocorrect).

Re: Open Problems in Robotics

#218

Post by jackhack »

bluGill wrote: post #206
robotresearcher wrote: post #147

Earlier quoted context omitted.

> Two robotic hands that can work in co-ordination just as human hands do. Extremely difficult to solve, but money is there. Unless real human hands continue to cost less.

Human costs are complex. I used to work in a company that made tape robots. One customer did the math, hiring humans to change tapes was cheaper than buying a robot. Then he walked in one night and saw the kids in hockey gear, take the needed tape off the shelf, slap shot it to the "goalie" who put it in the drive. Which is why he was a customer even though human labor is cheaper.

as someone who has worked his share of minimum-wage jobs, this seems obvious. Perhaps it's a case of academics in their ivory towers not having sufficient exposure to the real world? That kid hired for $8/hr doesn't care about the success of the business. Their concern ends when the shift changes.

Re: Open Problems in Robotics

#219

Post by acoard »

christopher22 wrote: post #196
Animats wrote: post #66

- Motion planning: already discussed. - Multiaxis singularities: much less of a problem than it used to be. We don't need closed-form solutions any more; we have enough CPU power at the robot to deal with this. You need some additional constraint, like "minimize jerk" when you have too many degrees of freedom. - Simultaneous Location and Mapping. SLAM for short: Getting much better. Things which explore and return a…

Kinda feels like that old life pro tip. Want to know the answer to something? Don’t ask a question, but make a false statement. Every expert in the field will rush to correct you and give you the most up to date and relevant information.

>Cunningham's Law states "the best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer."[0]

Also I'd like to point out that you incorrectly stated it was an "old life pro tip" instead of an "internet law". Anyways, here is the correct information for you.

[0] https://meta.wikimedia.org/wiki/Cunningham%27s_Law

Re: Open Problems in Robotics

#220

Post by christopher22 »

acoard wrote: post #219
christopher22 wrote: post #196

Earlier quoted context omitted.

Kinda feels like that old life pro tip. Want to know the answer to something? Don’t ask a question, but make a false statement. Every expert in the field will rush to correct you and give you the most up to date and relevant information.

>Cunningham's Law states "the best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer."[0] Also I'd like to point out that you incorrectly stated it was an "old life pro tip" instead of an "internet law". Anyways, here is the correct information for you. [0] https://meta.wikimedia.org/wiki/Cunningham%27s_Law

Thanks for setting me straight. :)
Post reply on HN