[SIGGRAPH 2020] RoboCut: Hot-wire Cutting with Robot-controlled Flexible Rods
HTML-код
- Опубликовано: 4 окт 2024
- Accompanying video for the SIGGRAPH 2020 technical paper
"RoboCut: Hot-wire Cutting with Robot-controlled Flexible Rods"
by Simon Duenser, Roi Poranne, Bernhard Thomaszewski, and Stelian Coros.
ACM Transactions on Graphics 39, 4, Article 98 (July 2020).
Link to paper: n.ethz.ch/~sdu...
Computational Robotics Lab: crl.ethz.ch/
Simon Duenser:
n.ethz.ch/~sdu...
Roi Poranne:
inf.ethz.ch/pe...
/ roiification
/ roiporanne
Bernhard Thomaszewski:
www-labs.iro.um...
Stelian Coros:
crl.ethz.ch/cor...
Abstract:
Hot-wire cutting is a subtractive fabrication technique used to carve foam and similar materials. Conventional machines rely on straight wires and are thus limited to creating piecewise ruled surfaces. In this work, we propose a method that exploits a dual-arm robot setup to actively control the shape of a flexible, heated rod as it cuts through the material. While this setting offers great freedom of shape, using it effectively requires concurrent reasoning about three tightly coupled sub-problems: 1) modeling the way in which the shape of the rod and the surface it sweeps are governed by the robot’s motions; 2) approximating a target shape through a sequence of surfaces swept by the equilibrium shape of an elastic rod; and 3) generating collisionfree motion trajectories that lead the robot to create desired sweeps with the deformable tool. We present a computational framework for robotic hot wire cutting that addresses all three sub-problems in a unified manner. We evaluate our approach on a set of simulated results and physical artefacts generated with our robotic fabrication system. 
Наука
Researcher: We built a robot that can use a wire to cut shapes out of foam.
Me: sounds reasonable enough.
R: and it can have a slack in the wire and with physics-simulation predict its path through the block to get large smooth curved areas.
M: That is witchcraft, and heretic sorcery to boot!
R: Unfortunately we couldn't figure out how to spin the block so we have an assistant do that.
M: I can see why a spinning turntable could be an engineering nightmare.
hahaha
Maybe ABB just charges 10 grand for that extra axis control? :-P
I just commented this before reading it 😂
Damn it! why are spinning tables so hard to create?
cuz it has to be super durable to handle the weigh-.... oh wait
i might be willing to donate my old vinyl player to the cause in case the assistant turning the worktable wants to take a day off
You know the assistant turning the table is the same person doing everything else, right?
this could prove very useful in a situation where a simple shape needs to be cut to safely hold something when shipping it. sure beats making a mold and casting it.
Genius
What it doesn't beat (neither by time nor probably in terms of cost) are those foam bags which perferctly fit the shape of the goods once steam gets blown into them. The foam-base material inside them expands through the hot, humid air (while being kept in place by plastic bag around it) - and voila.
Time needed for that ~3 Minutes.
Not every robot ought to have a practical application, they can also be a showcase of engineering feats, like the precision of the mechanism and the ingenuity of the software.
I miss the Dr. Károly Zsolnai-Fehér introductory. :(
What a time to be alive
Hold on to your papers.
2 papers down the line
Pronounced Karoy Zholnai-Feheir
cant believe 2 minute papers died
the ultimate dnd terrain bot
As a mechanical engineering student, I never anticipated that I'll find a SIGGRAPH paper so interesting!!! Now I'm binge watching all SIGGRAPH presentations on youtube lol. Somehow I got this vid recommended (pro'lly cuz I watch some robotics stuff at times) and I'm very glad.
The turntable isn't hard, but what about removing the excess material? Then again, a sufficiently powerful fan could help :D
You actually don't need to remove excess material right away. The wire would cut through anyway. It's done in the video for the viewers' sake only.
@@roiification If you have an option to attach a air hose somewhere on the hand, that should be enough for foam like materials like this.. Otherwise not much needed unless you are cutting details.
They allready have robot arms, why not use them for that too
Or another arm with a suction tip.
Just spin it around fast and the excess material will fly off.
All of the math, literally all of it
Index And yet such horrible English. Why?
I could smell the styrofoam fume from here
I don't know how useful is this in real life, but it's amazing!
Thanks for releasing this video publically. Very informative and inspiring.
I can image this being very nice for automotive exterior prototyping
It might be, but 3d printing already exists and I'm not sure it fills that particular niche better.
I wonder if it could be wired up the same way as an EDM and make extremely precise metal 3d parts.
@@andrewf8366 Seems as if it is much faster than 3D printing something at this scale.
Or, minis playground :P, for dungeon or similar
@@TheSunriseAnimation That is what i was thinking about aswell. Im not knowledgable in the field but maybe for aerodynamic models aswell
ruff cutting before the blank going to the cnc-router
Mind ---> blown. Wow. I'm working on a 2-axis machine that can cut foam in 2D, but this just rocks the boat.
Cool. Any idea how the rotation was synced to the cutting motion? How does the robot know the angle?
WoW. Truly Amazing.
working in an architecture office, i can say, this is game changing
I don't know why this was in my recommended, however this is really neat! Just a question from a guy on the outside looking in, what stops you from attaching a motor to a lazy susan/ or whatever works best for holding the material so you don't have to physically align the substrate. Just seems like it would be more accurate as well as not as tedious. You could even have it "stop and wait" for you to remove the excess material. The math behind this is amazing.
Glad you liked in anyway!
Regarding automatic rotation of the material (this question seems to pop up quite frequently): You are of course right, for practical purposes the workpiece should absolutely be mounted onto an automatic turntable - or even a third robotic arm. However, so far this has been a pure research project, and focused on the computational challenges. Since adding more hardware didn't promise any profound new insights, we chose to go for the cheapest and quickest option.
Seems that option is "an ETH student". :-D
Would be interesting how to scale this approach down a bit (e.g. for less mobile applications, these robot arms look expensive!) and also take into account the actual physical properties of a heated rod (melting foam away slowly over time if it stays stationary).
@@simonduenser5446 "We chose to go for the cheapest and quickest option" Spoken like a true engineer.
@@Sukrim Could probably put the workpiece on an XYa (left/right, forward/back/perpendicular rotational axis) table, fix one end of the wire axially to another XYa table and the other to an XY table. Accomplishes the same effect albeit with less resolution for way less money using parts with good economy of scale because of the 3D printer and home CNC markets. Might even be able to use an off the shelf CNC controller if the software used here outputs g-code or something convertible to it.
Really cool stuff.
@@Teth47 Clever idea! But the robot arms look like 6dof arms; with 2 of them plus a turn table you need 13 degrees of freedom, whereas each XYa table only gives you 3. Not to say that it couldn't be done with more consumer-level electronics, but it would be a real challenge at a minimum. Avoiding collisions could also be challenging, since it looks like the arms frequently move around above 3 sides of the cube rather than just one side (so you need to be able to reach around the workpiece)
awesome work by the way!
what a time to be alive
Assuming hot wire can deal with high density foam, this process can be very beneficial for things like automotive prototypes. Custom car hulls are AFAIK often sculpted from foam and then covered with fiberglass for rigidity. This process looks absolutely amazing!
this is INCREDIBLE
I like the aesthetic it generates.
this method is so powerful, it opens up so many possibilities, like using a flexible saw to carve soft materials like wood or soft metals perhaps, to reduce the time it takes to make large industrial parts, or help automate the process of making low number very large parts, I don't know what I'm talking about but I see the possibilities
I don't see this being useful for wood or metals. It would require the line to stiffen and be a straight line. Which would limit the possible geometries and require much more effort than tradicional CNCs (Not to mention that the back-and-forth motion of a saw would make this project a entirely new challenge).
I can see this being potentially useful for lost foam casting processes. Although it probably can't handle holes and concave details very well.
Actually... the more I think about it the less I see it being used in practice for any type of manufacturing...
But it is a marvelous idea and can very well be used in many future applications that haven't been discovery yet. Researches like this is what makes technology evolve in such giant leaps.
@@deox4225 you are right, this technology per se may not be as useful, but something inspired by this, or the next one after that may be more amazing
I really hope we'll get to see an open-sourced variant of the software in the future, assuming the affordability of robotic arms is more feasible in the future. A 3D hotwire bot would be a cool thing to have, especially when certain shapes are outside the scope of 3D printing and CNC machining on that scale is out of the affordability range for any hobbyist.
And yet I can't even manage to keep a hot wire tight for more than a minute
Metals expand when they heat up, try tightening it up a bit more to compensate
I find it interesting that despite the advanced technology, the traditional method of making a coarse "rough draft" followed by finer details is still used. The same applies to expert artists and scooters
Imagine what the robot can do if it could also control the foam's rotation (in 3D) while it is cutting :o
Not much more, it already has 6 DoF with the wire.
The infinite molds this could produce! so many parts - so many ideas! 8D
Ahhh! I thought this was animated until the humans arrived in frame and threw off my perception of reality.
Robot: then the humans started playing with the cutoffs and not the model. Typical.
First thing I thought: Large chess pieces! Too bad it doesn't rotate.
Very good.
Always suspicious when a shot is cut (end of sphere cut shot at 3:40 ).
Great work regardless, thanks!
I was thinking the same thing
Just finished my Bachelor's and had decided to switch my field from robotics/mechatronics towards computer science and do more of the AI/ML stuff related to robotics rather than pursue the mechanics. This video and related research projects just validate my motivation even more. Thank you for this.
Maybe it's a coincidence but my Bachelor's thesis also involved trying to replicate an ETHZ paper for a different robot.
this is BEAUTIFUL ,, imagine if the robotic hands could manipulate where they grip the wire for further settings , or even , "change" tools xD
they could be like michael angelo
Twelve degrees of freedom, plus one intern, to achieve three axis machining. Well done, academics. Well done.
Congratulations on your idea! Great job!!!
This is incredible
Adding one or two additional hands to rotate the table would be a great idea to remove cut parts. It would add also a great amount of futurism to tye design. Two arms cut, one or two holds the table and rotates it. Everything is moving.
Wow! Very impressive.
🤯😍😍😍😍😍
This is incredible. Thank you for sharing.
Absolutely fascinating. I don’t know why this was recommended to me, but this is quite cool. “We will watch your career with great interest”
Subscriber earned.
This is the next 3d printing
Great job guys!
Amazing work.
Interesting implications for quick-scale mockups. People have been talking about 3d printing for years, yet I could imagine this tech could be applied to cut out several different shapes in the time it takes for one 3d print! Different scales of detail to be sure.
Getting horrible future visions of bald business suit wearing industrial robots rolling around with superheated garrotes, lobbing human heads off in a blink of an eye with incredible precision.
Super cool work! And it addresses a real need
😂 Good one.
shut up, we do need it.
I took a class in arts academia where they teach about integrating new technologies into sculpture and do you have any idea how damn long it takes a conventional cnc arm to carve a block that size or larger? a whole day if you're lucky, I would've killed to get one of those in the lab.
to be fair the arm we have also carves marble at the same speed, but who the hell can afford marble when foam is so cheap and convenient
@@npc6817 I does address a real need. I literally said as much.
Think about large scale architectural forms, many are made of foam. This can be a useful way to prototype, and maybe even to manufacture those custom pieces.
@@AlanMelling sorry, I thougt it was sarcasm.
I mean, that's what we used the carving robot for, we made those large foam prototypes since it was impossible to build them in marble, but even when we used pre-shaped blocks to ease the work on the robot it'd still take ages, this one could've easily done it in half the time
@@npc6817 no worries! That sounds like fun work! Seeing your ideas made manifest in reality is fulfilling.
Phenomenal!
WOW great job of engineering
That’s pretty incredible! I assume the slow speed is chosen such that the cutting forces are insufficient to deform the wire, because I’ve seen hot wire cutters move much faster than that. Calculating the static curve of the wire looks like an interesting analytical problem, but adding dynamically changing cutting force to it would likely make it a computational problem, and not necessarily a fun one. Adapting the rough cutting path to the form of the solid is a problem I’d have no clue how to even attempt, good job with getting good results out of that!
Great work! How long does the toolpath computation take? And does the user have any control over the process or is it all automated? Loving it!
Regregards from Poland Mr Tomaszewski!
Nicely done!
So sick. I'd love to have this to make fiberglass negatives for carbon layup without machining insulation foam and making a huge mess. Cut foam > fiberglass > Prepreg
i think the best piece is the last one, i'm sure this robot excels in doing curves that are already in the piece to be cut. while the rabbit and other animal forms are cool, they are really rough, they could be a first step before going under a cnc router and get a better finish
This is quite amazing indeed!!
Having a turning table would also allow even more Shapes. Imagine the Workpiece turning with differing speeds and directions, while the wire is cutting. You might be able to cut out stuff like double helixes.
Nice, i suppose this is faster than 3d printing high volume shape prototypes. The multi segment especially.
and way faster than milling
Incredible. Super awesome !
Is this an 'open source' project by chance? With code released as well?
Thank you for your interest.
Unfortunately, at this time the code is not open source. We are currently still evaluating the best options to facilitate widespread utilization of the technology.
Meaning they worked hard on it and they need to get paid.
@@pizdaxyu You know what stifles innovation?
Not getting paid :/
@@pizdaxyu Personal enrichment? Are you a communist? Let me kindly ask, what do you do for a living, comrade?
@@pizdaxyu most of the time the school you develop at has rights to what you create in some form. stop acting like the world should give you everything for free.
Absolutely amazing
this is veeeery interesting
Really nice idea and project😁
This would be very useful for the production of lost-foam casting for one-off projects. As in, if you don't plan to produce more than one or two of anything, then this would be ideal.
I would love to have that to make composite sandwich structures :o ! The path prediction is impressive!
Cool. Any idea how the rotation was synced to the cutting motion? How does the robot know the angle?
My guess is that they put together a series of cuts manually in the simulation, exported the arm movements from that and re-played it. So the angles were already known, possibly set arbitrarily by a human.
That's really awesome
This would be really interesting to create hard-to-make curved formwork for concrete or plaster molds. With some kind of interior coating, i think itd be interesting to make reusable molds.
to be honest, What I can see is the future in tumor resection in optimizing area of resection ( like Glioblastoma Multiforme ). Many of the time, the tumor is difficult to resect due to margin involving functional area, and the margin is difficult to judge . In radiotherpay of cancer, gradient field optiziation is used currently. but curvature optizimation may still be a difficult topic in the RadioTherapy treatement
Now if they added an automatic turntable, it could act as a sort of lathe.
Very cool
This is freakin awesome!
Wowww Flexible body simulation !
Great work.
*the cnc mills are evolving*
The robot could do 2 cuts per turning, one on the near side & another on the far side (assuming the cutting direction has no impact on the cutting result?).
( eg. when milling it can be important to cut in the correct direction, because it removes material differently depending on material movement relative to the tool rotation )
An example of this is actually shown in the video
Very cool! Keep up the good work!
Have you considered using this for fiber glass form making? I'd imagine you could do some fairly large scale automotive paneling or boat hulls from 3d models instead of clay carvings they do now
Very cool, this could be automated even more by adding rotating table and something to remove excess foam
Removing the cut-off material shouldn't be that hard either. While I get the generall idea behind "use a fan to blow it off", the drawback is a possible canting of the cut-off material with the remaining, so it won't be blown away.
Two ideas come to my mind:
1) implement a vacuum-nozzle (no high suction force required because of the lightweight foam materials) on both robot arms that is attached e.g. on the opposite side of the hot-wire mounting points and use the nearer/better suited arm to pick out the cut off pieces.
2) Use a third arm with the method outlined above. Sounds more expensive, but might be easier to implement.
Both methods would allow to either push the waste material away - or, if that is not possible, pull it out.
Amazing. Thank you!
Really nice!
Amazing!
Very interesting
Incredible work!
Humans are too smart. It's insane.
Have you thought of trying to make hollow pieces? where the wire is bended to a curve and rotated inside, making a vase type sculpture?
That's genius! Hot damn!
cos players could make some crazy looking costumes with this.
🔥🙏
Make a model plane!
Or like a glider or something. Point is, something that actually flies
Best work quickly, 3D printing is becoming far more viable for large scale prototyping
I can make you a revolving platform for you xD
Incredibly interesting softwere. I don't even know what kind of algorith you have used.
I wonder if the same principle can be done with wire EDM
Electrical Discharge Machining?
The erosion rate of the wire/tool in EDM is fairly high, so you would have to feed a spool of the cutting tool bar through the robotic arms while cutting, or at least to replace the tool every so often to get very far through a metallic workpiece. Interesting idea.
How do you predict the way the wire is going to bend?
The temperature is not going to be constant, because the part of the wire which is doing the cutting loses more heat, so the mechanical properties change along the wire. Furthermore you have to consider the force applied by the foam on the wire.
I would be curious to know how you approached such a complex problem
This is very impressive, but what would the practical advantages of this method be? Compared to regular CNC foam cutting.
speed, curves and surface smootheness?
Why did you use the double-actuator ABB instead of a pair of robots with a common controller? I've used Fanuc, Denso, and Epson units that have a slave/master controller system to do stuff that appears superficially similar (hand-offs, crash avoidance, team lifts) and they can definitely achieve the endpoint precision required, but I'd be curious if that ABB 2-in-1 has tighter coordination while in motion.
awesome!
Was it just for the challenge or is there a reason why this approach would be preferable over just using a single arm and a pointy hot-end to carve it in a more detailed manner?
I sometimes wonder why some smart people try to automate things artists usually make.
Still trying to figure out how this would replace consumer foam products where economies of scale make molds far more effective than this. There might be a niche application in making a positive for some smaller molds for tool creation but even then you're looking at post processing either using a mill or an artist. Either way it'd be interesting to learn more about the algorithms than see more of this.
You need a piercing heated rod on a third arm and then you could subtract from under arches etc
give it a meatfork
Interesting even for a layperson like me, as you can't touch a straight wire following a concave path even with 3 axis rotation and only have it contact at the controlled point but this can. If you could make that more rigid wire sacrificial or a guide for a sacrificial wire so it can be changed as it wears during a cut, this could lead to a new generation of edm machines capable of even more complex eroded shapes without custom electrodes.
The wire doesn't wear.
@@nagualdesign in the current machine yes, but in a wire edm (electro discharge machine) the wire erodes away continually and has to be fed off a spool while the machine runs. Most wire edm's already control the top and bottom wire guides giving 5 axis for mould relief cuts etc and some newer machines incorporate rotary axis for the workpiece giving even more flexibility on machined shape, but there are limitations on what forms you can cut with one because the cutting path cannot describe undercut for example.