Robotics

Didn't see one in the catalog. Are any lainons into robotics? I want to get into it but I'm not sure where to start. Sorry if this is the wrong board for this or this is too short an OP.

OP here, to add more discussion to the thread, I don't have much experience with coding or engineering, so I don't know what's possible. What I have been daydreaming about is a little robot that can follow me around, accept voice commands, and maybe have some self defense capability. So I would probably need to make a robotic turret that can track and engage targets (taser or something else non-lethal), then slap that onto a cheap robotics platform (probably a cheap 4 or 6 wheeled aluminum chassis with direct drive motors) and power it using a salvaged computer and cheap off the shelf sensors. How hard to put together would something like this be if I don't know much more than scripting, or basic DIY? How hard would it be to unify these functions into a single package? Would software packages like ROS make this easy enough to do, or do I need to know how to code drivers and other modules myself? Anyone ever done anything like this before?

>>2214 I am really into programming and studying electrical engineering and I have been wanting to build some kind of robot for a while now, but I have no idea what kind. Maybe some kind of robotic arm, walking legs or robot cat would be cool. I really like to look at the boston dynamics stuff and things like a stationary head on a moving robot (chicken mode) and other really cool movement stuff fascinates me. This stuff actually isn't really that hard to imploment, you just need motors which can execute a lot of force, and have good feedback/resolution, and that stuff gets expensive pretty quickly. I have been thinking about maybe making a servo out of a belt, a brushed or brusless motor and a motentiometer/rotary encoder. But this also gets expensive really quickly, with something like a brushless motor + esc quickly costing €30 for a cheap one. The cheapest way would be cheap servos but this would just get you a soykafty choppy robot toy.

>>2214 Did a little in my 3D printing days. Main limiting cost is high quality motors for many applications, which is like kicking dead whales down the beach. Buying in bulk didn't help either.

>>2214 Motors and motor drivers are the most expensive part of building robots Even the chassis can be relatively inexpensive if you 3d print, but motors are not. Your best bet would be to modify an off the shelf RC car, they're inexpensive, dime a dozen, and they have everything you'll need, maybe a small gearbox to scale up or scale down, motors, a great body to work with, battery holders etc

>>2231 That and drones

>>2214 I am a mechatronics engineer for a startup. Robotics is just one aspect of what I do, but it is among my favorites. Ask me anything. I still am working on these actuators called Twisted Coiled Polymer Actuators. They have the potential to bring high power, high precision, low cost "muscles" to robotics applications. My research centers on improving their work efficieny and overcoming the limitations of their high operating temperatures. https://www.sciencefriday.com/articles/how-to-make-an-artificial-muscle-out-of-fishing-line/ https://www.utdallas.edu/news/2014/2/21-28701_Researchers-Create-Powerful-Muscles-From-Fishing-L_story-wide.html neither of the above is my work, but it is roughly where I started.

>>2239 I've run across those articles before when doing research very similar to yours. There are so damn many different types of artificial muscle it's hard for an aspiring robotics engineer to know where to start if they decide to use that instead of electric motors, hydraulics, etc. I recently had the thought to use chemically-actuated muscles, but that would necessitate a pump system to feed them the chemicals they need to move, so I'm really not sure if I should bother with that. There isn't enough of a basis for me to go on here since very few systems have ever used this. There should be a comprehensive database of artificial muscle types that lists them by force output, mechanical properties, etc. The science probably isn't advanced enough for that yet, unfortunately.

>>2215 >How hard to put together would something like this be if I don't know much more than scripting, or basic DIY? For each part: >So I would probably need to make a robotic turret that can track and engage targets (taser or something else non-lethal) There are open source implementations for this exact thing, via face tracking and the like. An entertaining example is https://hooktube.com/watch?v=d5d02U5YYfk >accept voice commands shouldn't be too hard to translate voice to command, with the voice to text APIs available. You'll need a simple grammar so that you can parse the commands, but as long as your command structure isn't too complex it shouldn't be bad. >that can follow me around also open source implementations for this. object tracking + movement. The hardest part for you would probably be wiring everything up correctly and not accidentally frying your circuits, and I don't want to imply that this would be a trivial project by any means, but it should be doable.

>>2239 How feasible is hydraulic actuation for DIY systems? I like the power density, but I can't think of many smaller platforms which use it, which isn't the greatest sign. Also, speaking of thermal actuators, what are your thoughts on those "popcorn actuators" that went viral recently. Think they are good proof of concept or just a fun diversion? >>2240 >I recently had the thought to use chemically-actuated muscles, but that would necessitate a pump system to feed them the chemicals they need to move, so I'm really not sure if I should bother with that. I may be wrong, but I think that many of the better preforming robotics platforms have pumped coolant systems. Depending on what you ultimately go with, you might have to deal with pump systems anyway for heat management. Going with pumped chemicals could kill two birds with one stone. That's just baseless speculation though.

>>2241 Sweet. Thanks for the reply and the amusing link lain. I'll start looking around for a decent chassis kit or see if I have a set of good enough motors and wheels to work as a prototype and go from there.

>>2242 >Depending on what you ultimately go with, you might have to deal with pump systems anyway for heat management. Going with pumped chemicals could kill two birds with one stone. I'd been wondering about how I was going to cool this thing anyway, but a mass of hoses can make the whole thing overly complicated and failure-prone, so I'd like to find an alternative pump type if nothing else. Artificial muscles can have as many different power sources as you can imagine, ranging from chemicals to electricity to heat and many more, and a database for them would of course list what it needs to operate. But what I'm the most concerned about right now is pinning down the operating mechanism of the muscles. You can have a tanker truck full of gasoline, but it's worthless without an engine unless there's a pack of marshmallows that urgently needs to be roasted. I'd like something very similar to how natural muscle works because you can then import medical knowledge of biological muscle instead of having to work out the dynamics of a completely different system. I haven't heard of >>2239's twisted coiled polymers until now, so I have to ask how similar they are to natural muscle. Looks like I have some research to do about this, but it'd be a major help for him to tell us about how they work or suggest some alternative paths for us to follow if they turn out to not be suitable for a certain project.

>>2239 That's cool, someone else who works with twisted polymer muscles! I have had some great success with those. Cheap, strong and efficient. I am actually working some more today and can post some pictures. >>2215 As for OP I would suggest you start learning engineering and programming if you are going to work on robots. Start with something simple like free code camp and The New Bostons physics tutorials. Then go try ROS, write some robot AI packages and pickup some robot books.

>>2274 Those twisted polymer muscles sound kinda cool, but as far as I understand they work by heating them up. But isn't this a super slow process, I can't imagine using this in a robot.

>>2276 It depends on the thickness of the wire and the heating element used. I have 8lb test fishing line that contracts and releases in about a second. But my 200lb test has a duty cycle of about a minute. But the thicker the diameter of muscle the more strength. There are also hydrocordial versions of these muscles which are annealed shaped like springs. They can contact up to 40% in my experience but their strength is much weaker than the traditional for. UT Dallas has a hybrid design I call a "helix' that combines both methods which I have seen great success in. Picture related.

>>2242 It really depends of scale. It is not too difficult to get a working fluid hot enough to drive the actuators, and building sheaths around the muscles for the fluid to flow through should not be too complicated or expensive either... Actually, the smaller you make the TCPA the faster it can operate in passive air cooling operation. Smaller muscle = smaller thermal mass, thus less to cool off after reaching operating temperature. There has also been some work looking into using LLDPE or other low anisotropy fibers to improve work efficiency at the cost of power density and tensile strength. You might check that out if you want to go smaller, presumably at lower loads. Current models of the TCPA suggest that their thermally actuated behavior persists down to the scale of micrometer diameter filaments, possibly down to the molecular scale. So small is actually a more viable use case than large in many ways. As far as the popcorn actuators go, I find them interesting, but I do not know enough about them to comment intelligently.

>>2276 https://www.nature.com/articles/srep36358 Thin LLDPE can achieve very short cycle times due to its low operating temperature, increased work efficiency and reduced specific heat compared to nylon or other fibers. It is the most work efficient material used in this application to date, even more so than CNT or paraffin impregnated nylon monofilament.

>>2274 It is a pleasure to meet you fellow explorer. :) I would post pictures of my work, but I am about to publish a paper in a month or two, so I cant really afford to give away all my secrets lol.

>>2274 But, as a special present, I can post some videos of the machines we have built and maybe a talk or two... As bad as they might be... https://www.youtube.com/watch?v=I7akKDGRk_A https://www.youtube.com/watch?v=YJzdtgj7G38&t=0s I guess this blows the whole anonymous thing huh? :P

>>2248 The exact mechanism of the muscles is still not fully categorized. They are still being modeled and there are ongoing debates about how exactly twisting a monofilament until twist becomes writhe amplifies the effects of thermal expansion, and reverses the expected direction due to such expansion. It seems like there is likely more at work than mere thermal expansion to me, but my research work has been mostly focused on production methods and improving work efficiency to make market viable products. WE have started characterizing the fibers we produced with our own custom materials, but there will some months yet until we start really digging into what makes them tick. We intend to put them through a rigorous battery of tests including Xray diffraction, optical diffraction, Electron microscopy, and some others. Maybe we can learn a bit more about what molecular activity is going on then. As far as finding other directions for your project, if you really need muscles that behave like human muscle and have comparable performance, you are probably soykaf out of luck. You could grow some biological muscle in a dish and provide it with artificial vascularization maybe. I can say with some perhaps misplaced confidence that the muscles we are producing will have work efficiency on the order of 10%, contractions approaching 25% of uncontracted length, and maximum loads up to 9 MPa. But even then, we are at half of what the least efficient biological muslces are recorded to be capable of in terms of work efficiency. Much stronger, and with longer contraction lengths though, also, no fatigue or oxidation to be concerned with. You could go a few different ways if you dont need exact human replicas;NiTi wire is widely studied and well understood, pneumatic or hydraulic soft actuators, various shape memory polymers, piezoelectric actuators, electroactive polymers. Really there are too many possibilities to discuss without knowing more about the specifications for your particular application. If you want me to offer more useful information for your specific thing, I am going to need some very specific performance parameters and a detailed list of desired operational limits.

>>2286 That machine looks great, how fast is your manufacturing process? I will keep my eyes peeled for your publication, I keep up on all related white papers. I am more of an advocate, hoping to document and bring attention to material actuators in the hopes that more funding and research will be done. Mostly so that someday I can use it for robotics I admit. I'm the same spirit of shared exploration here is my Hackaday project, OP may also find this useful. If not HBS labs on YouTube has a video about manufacturing TCP. https://hackaday.io/project/157004-material-linear-actuator-for-robotics

>>2292 We have gotten it dialed in to being able to make about one 5 cm muscle every 10 seconds. We could have many machines in parallel if we wanted to go faster. Something I have been mulling over, that you might find interesting is the possibility of use TCPA as a data recording medium.... or perhaps to create true artificial DNA. https://en.wikipedia.org/wiki/Wire_recording imagine if the wire were twisted the same way the TCPA are, and could be untwisted in a localized area reversibly for reading. One could get enormous data densities with relatively easy to produce spot sizes on the surface of the TCPA. I imagine an optical method of reading and writing onto the TCPA as in optical tape... https://en.wikipedia.org/wiki/Optical_tape As far as artificial DNA goes, DNA is already a polymer material, the knotting of TCPA goes to show that the twisting of DNA into nucleated knots is not unique to DNA but a common part of the behavior of polymers in general. I have done some experiments with flattened nylon and trilene fishing line and found that they nucleate and coil/uncoil very similarly to biological DNA. I am still working on the implications of this behavior, perhaps it will stimulate your imagination like it did mine :) if you want to get in touch, send me an email at the address I have provided lynnmt@firemail.cc I cant promise I will always have time to chat, but I do like to be in touch with people in the field.

>>2292 You can also get in touch with my PI, Dr. Hector Ochoa ochoah@sfasu.edu He is pretty busy these days, but I am sure he would be happy to talk to you about the research we are doing in his spare time. No promises though.

>>2294 Great, I'll send you fellas an email sometime with some inquiry. Glad to see lots of progress and new ideas being thrown around in this area. Electric motors have over a hundred years of innovation SO I wonder what smart materials will look like with the same amount of time.

>>2294 >Dr. Hector Ochoa why does that name sound familiar... did he recently publish some groundbreaking paper? I feel like I have read some of them in the past but cannot place them > I have done some experiments with flattened nylon and trilene fishing line and found that they nucleate and coil/uncoil very similarly to biological DNA. I am still working on the implications of this behavior, perhaps it will stimulate your imagination like it did mine :) isnt there an algorithm or something that can determine the method in which DNA/polymers will coil uncoil? as in there is a mathematical direction imparted by some force or another that could be used to predict polymer behavior, essentially allowing for directed coiling and uncoiling much like an epigenetic driver? forgive me if I have misspoken, I actually know know nothing about biology or engineering

>>2322 to add on to that, could one potentially use a polymer coil/uncoil system to simulate some form of logic gate?

>>2322 There are many such algorithms designed to measure the effect of many different forces acting on every part of a biomolecule. Each is suited to a different situation. Physics-based approaches have a great deal more computational expense involved but deliver more accurate and precise results, while statistical approaches have less accuracy but run much faster.

>>2322 IDK but you need to slow down with the word salad. Feel free to play around with twisting flattened fishing line and see what happens. :P

>>2327 Thought I would come back and post about the arm I made last month powered by twisted nylon pairs, in order to get motivated for a new rig I'm going to start this weekend.

>>2515

>>2224 You CAN safeway it though, make one with soykaf servos as a prototype sorta deal. I mean, maybe, MAYBE, someone on the market is gonna be like I NEED THIS and you get funding for a real one

>>2515 Bro, is that the method with keeping a damn fan near it, or does it heat up be electricity? I mean, i can't really tell and i kinda need to get into the synth muscle stuff

>>2516 This guy gets it Also, would it be useless labour to combine pairs into musclegroups with something like a stretchy polimer as a casing to simulate muscles completely, kinda sorta like pic related

>>2577 It's electrically heated by wrapping the nylon with Nichrome wire and passing 16vs through it. Large strands have a duty cycle of about a minute, but lift a pound. Smaller strands have a twenty second cycle but can lift less than half. >>2578 Well coating it in something makes it unable to contract, I experienced this when using conform coating during an experiment. I'll make a detailed video of how I make it if someone posts a picture of Felix Argyle they drew themselves. Additionally I have several small robots AMA.

>>2216 ButtplugBot has a nice form factor, does it work?

>>2803 Um actually, it's a sentient jap fleshlight

>>3413 With that attitude, maybe. What's to say it's not both?

>>2226 This The biggest cost is motors. Any motor that can rival a human in terms of torque gets quickly very expensive, and you need a whole different subsystem of power electronics just to control this thing. Bldc motor are great, but they require a controller to drive them, stepper motors for fine control are also complex, and the cost vs torque graph is exponential. Servo motors, the ones that can do any kind of appreciable work are very expensive. Everything else is peanuts compared motors

Arduino is a good place to start because it's native 5V and you will want that for the motors. I2C is the USB of serial buses. There's vcc, gnd, clock, and data to hook up. Assuming each sensor has a different id, you simply daisy chain them together. If not, you can use a multiplexer. If you need higher voltage stuff, you can use relays. It is also possible and sometimes necessary to use separate power. Simply grab a linear voltage regulator so the output is at 5V or whatever the sensor or motor requires and the input power can be whatever the regulator is speced for. There are loads of Arduino code examples for every sensor under the sun. They make it far to easy to get your foot in the door.

i'm robo-user, more of control science guy, but I'll be happy to help

>>3443 I know what you mean and don't mean to disparage the rest of your post but: >the USB of serial buses

>>2578 Wouldn't it make more sense to coat/electrically isolate the electric wires and then flow water around the whole muscle structure (e.g. "arm") to cool it? (Ideally the coating would be the polymer itself, I think.)

>>3508 Okay that was a retarded thing what I said. It's shares similarities.

>>2214 If you want to get into robotics, I recommend starting by building a 3D Printer. It is both a great learning experience and a awesome tool to have on hand for the actual robot building. You don't need to actually design your own, there are tons of awesome open source printers out there. Alternatively, if you don't want to source all the parts yourself, the cheap chinese kits are a very good learning experience - they keep breaking so you are forced to learn to fix them. Or you spend the money on a prusa i3, they are reliable but cost more. In any case, order as a kit. I am only building small and soykafty robots and do robot battles vs a friend of mine but my 3D Printer is the most used tool in my lab. You design a robot in CAD (or modify existing files if you are lazy), print the parts and slap your electronics in there. And even if you build proper machines from metal, 3D printing parts to test your schematics can be very useful.

>>2214 I wish someone would make simple ball jointed robot girls already with giant Megaman (X) tier universe legs. They can't balance, then let them Roll around and use basic chatterbox AI that's 90s tier. Infrared lasers to keep from bumping into stuff and just be. No one cares enough to do it. Yes it's hard, but we have NOTHING so far.

>>4074 Build one

>>2217 This is better w/ sound. >>4074 I'd say visit /robowaifu/ if you dream of having a robo wife, but the original haven is gone now. So you should try find wherever their refugees ended up if you wanted to see how their attempts at robotic gfs are coming along. There were quite a few good projects from memory with collaboration from the /clang/ autists.

>>2217 Decided to finally look up the source: https://web.archive.org/web/20091228051318/http://clockwork.shikisokuzekuu.net/HINA.html http://lannyland.blogspot.com/2009/03/robot-of-day-hina-cutest-mini-robot.html http://correderajorge.github.io/Cool-Robots-Hina-ver.RF09/

>>3439 If I wanted to build a simple SCARA like tower-bot, able to pick and place 20kg, say 0.5m or even 1m away from the base, and do it as quickly as a human or within 5s. How big motors would that have to be? Would servomotors even be able to do it, without having to resort to hydraulics?

I'm a industrial robotics programmer in the automotive sector, ama

>>5729 It must be really monotonous job. Isn't it?

>>5730 to be honest it is very fun when doing installation for a new line. Buggfixing is horrible, but i can't imagine any programming job would have engaging buggfixing. Day to day is usually quite fun, I can spend alot of time hanging out with the "physical" maintenance team, as they are the ones who has to make sure my/our code actually builds cars

Has anyone had any luck implementing the methods described in this course http://underactuated.csail.mit.edu/index.html in a real world system?

>>5729 Europe, murica or Asia?

>>5729 What sort of robots do you work on/with? Are on hardware or software? How much stuff can those robots do? Is there any decision making involved or is it all just repetitive movements.

>>5731 Any tips for beginner who started learning on his own?

I want to make myself a robot as a winter project from an old eeepc, bunch of webcams and serial port, plus aluminum frames and some wheels. Il keep ya posted how it goes if i can find this thread again.

anybody bought stuff from youbionic? Tempted to bite the bullet and go for the arms

>>5193 >I'd say visit /robowaifu/ if you dream of having a robo wife, but the original haven is gone now. So you should try find wherever their refugees ended up if you wanted to see how their attempts at robotic gfs are coming along. https://alogs.theguntretort.com/robowaifu/ The board moved here, it's pretty nice.

>>7899 guntretort is kill, robowaifu gang is now located https://alogs.space/robowaifu/

Hi Anons, I run this site and have this online paintball marker: https://remocon.tv/5fa8b5d8f5101758e9559665 If any other anons want their robots to be interent-controllable then feel free to find us through links on that site and we're happy to help, there is a free API that's easy to use. OP in my opinion what you're describing is very possible. For the parts of it that don't seem possible I think people are probably building out the tools now such that, once you've made decent progress on your robot and are half finished it, those tools will be ready and you will be able to use them on your robot. If you go with a wheeled one it would probably be done faster, but a legged one definitely looks cooler right now. Starting off I would say to just control motor drivers directly and to figure out ROS or high level stuff later, because it's far less overwhelming and you can do pretty much everything with just low level knowledge. Use ROS when it's useful to solve a problem, but there's no need to start with it. What I would say is first build the actual functioning parts and then structure and make it work, then make it autonomous after that. So build your turret, your robot, control it with basic signals, then you can get to the autonomous part. Start with the hard parts and build the easy parts after. Program your computer to move motors, then build your turret around those motors, then figure out the more complicated parts after. The general structure you described is good, an aluminium chassis, motors, sensors, like you said. I am about to buy a Rock64 or RockPi as an alternative to the Raspberry Pi, there are other single board computers out there, even small Intel NUCs or other computers work, get something with GPIO pins though. After you have your robot and you can control it, I would say look into OpenCV, computer vision. Maybe you would use YOLO or whatever the top computer vision models are at the time that you use them. That way once you already have your motor control, you can program functions like 'if <target in image>, get <location of target>, move turret X amount', 'if <target> is in location <X,Y>, shoot', or program a self-driving thing to follow you around. Computer Vision as the sensing, and then your programming to drive the motor controllers appropriately. Again this field is advancing so quickly that once you have your robot built the tools to abstract the controls might already be widely available, so you can just plug and play. I notice that your post is from a few years ago but it's still doable and this applies to anyone reading who also wants to get into it. Another bit of advice I would say is to get a 3D printer, it saved me a lot of hassle, although my first several 'robot'-like devices didn't use one. Also don't stop if you hit a roadblock, try to find other ways of accomplishing the same thing, sometimes later on you will come back and realize how you could have done it, but with the right attitude you would have succeeded anyway because again you would have found another way to do it. A lot of people stop when they hit those roadblocks. >>2224 You can buy one of those Pi-car kits and instead of a Raspberry Pi use some other single board computer (e.g. some Rockchip one or one of the others). Also if what you are building is small then I don't disagree with the idea of buying a bunch of those MG996R servos and using those. Those servos are cheap if you look in the right places and you can buy a bunch at once to make a robot spider or robotic arm. If you know of larger or better servos that are a similar function/cost ratio feel free to let me know. If it can lift a lot more I can pay more.

>>2224 Check out the website that the anon is referencing: >>7895 Even that very impressive robot arm, if you zoom in on the GIFs and videos, just uses cheap MG995 servos by the looks of it. I commend them for it, they get the job done, and look how good those robots are that use them. >>7895 I haven't bought anything from them but just wanted to say that it's very cool what these random independent roboticists have made on their own. They are very underappreciated compared to the handful of companies that get most of the media attention.

>>2240 >>2240 >"there are very few systems that have ever used this" Every Animal Ever has used this system. Most of them dont even know what electricity is. I haven't personally looked into it but i would be willing to bet that similar setups could be made.

>>10299 Obviously they were referring to artificial systems lain. Just because something evolved doesn't mean it's easy or even feasible to build.

Bump. Any guide how to start? What knowledge i need to start?

Any thoughts on HASEL actuators, lains?

Presented without comment. Atlas.

>>11529 New atlas is holy soykaf, just something amazing. Why's it only Boston Dynamics that can make bipeds that don't fall over + walk fast?

>>2214 BUMP I get how learn electronics but what about mech part? How to design them? What material use? Need books,courses

>>11529 I wish I could get a job at Boston Dynamics.