I read Robotics: A Very Short Introduction by Alan Winfield and it mentioned building an e-puck, but that still leaves a big hole before building quadrupeds
How do I build one of these?
I read Robotics: A Very Short Introduction by Alan Winfield and it mentioned building an e-puck, but that still leaves a big hole before building quadrupeds
did you build your e-puck yet?
I'm trying to build it virtually first, the book mentioned a lot how everything has to be done first virtually. Still figuring out Webots and Gazebo simulator but trying to think of a timeline of basic robots to build up to something like spot. I got a basic idea of e-puck, then figure out how a simple arm works by studying the PAROL6 is, an open source robot arm.
https://github.com/PCrnjak/PAROL6-Desktop-robot-arm
and then try to recreate virtually the hexapod
https://github.com/MakeYourPet/hexapod
The idea is to try and recreate these robots virtually and then see how I can combine the ideas, the epuck seems to be for understanding the meat and potatoes with sensors, the arm for the motors, the hexapods for legs but this feels really naive. I been trying to stick to everything open source but I feel like there must be a better way. I feel like I'm missing a bigger picture and so I been picking up a lot of physics books especially statics related but again idk looking for ideas. My background is in CS and I've played with plenty of microcontrollers but the actual robotics is new to me
>I feel like I'm missing a bigger picture and so I been picking up a lot of physics books especially statics related
sounds like you're on the right track, finish up that book then get back to us
physics and static and dynamic engineering are two very different things. you want static and dynamic engineering. next you'll want materials science engineering e.g. how shit fails, sheer, rotational load, structure. You don't ever need a physics book, you're doing actual shit (all of which is derived from physics).
>what is forward kinematics
>what is inverse kinematics
>what is a pid loop
go back to school you missed all of the required classes...
Even if you had the working, physical dogbot in you possession, you would be unable to write the software to make it work before your life force expires.
these kind of comments are always in bad faith. if I were doing everything alone from scratch, I would make a thread about it. As soon as I know enough to get started, at a later date I'll post a github repo with a discord , find other people to work at the local marker space. It doesn't matter if I can't write the softwsre entirely myself, I just need to know enough yo coordinate with others to do it and if boston dynamics can do so can others
Fourth post, best post.
There's already ripoff of Boston dynamics dog.
>Do it yourself
>Someone else already did it
>I want to learn from works of others but I don't want to learn from works of others
troll thread
It wasn’t in bad faith, I’ve seen lots of hardware projects abandon or ruined because they underestimate the gravity (funny physics calculation joke there) of the software situation.
> all of which is derived from physics
It’s the opposite. Physics was derived from observations & “doing shit”
Originally, anyway.
>Physics was derived from observations & “doing shit”
A nuance which has been completely lost, thanks to the grift that is university publishing. Always study applied first.
>Always study applied first.
but how do you do this for cheap? it costs infinitely less to do theory and run things in simulator than actually having to buy materials
wow anon, I didn't know you could substitute a multi-phase model predictive control algorithm with discord, a github repo and the local maker space!
>other people to work at the local marker space
I doubt your local gay bar will be of help
>local gay bar
what's the alternative? not everyone has a lathe in their garage
The best project by far you can do is outfitting your shop which then serves you for the rest of your life and easily saves much more than it costs. I learned late but putting tools before toys later yields many toys while making the rest of life comfy and more affordable. I DIY basically everything I need to maintain, modify or make as anyone can if persistent and determined.
An impressive amount of equipment can fit in a one-bay garage especially if you use high areas for storage like shelving near the ceiling. The rolling base you use to move your lathe and mill (if you want those) can be left attached so long as the machine is levelled at its point of use. Mills can sit on two pieces of large angle or box tubing so a pallet jack can go beneath.
The earlier you begin the better and making everything easily mobile is cheap while allowing easy reconfig and cleanup. Your shop will be vastly more useful than a single toy because it can make many toys.
You don't need to machine everything yourself either. The professionals (wiser than DIYers in all related ways) sure don't and even fully equipped machine shops outsource shit they dislike doing (like Blanchard grinding) or is not cost-effective to do in-house (laser and waterjet cutting).
One way to get machine time is taking a CNC course at your local community college then being helpful (volunteering is fricking gold) in return for permission to use the machines. Schools like cool student projects and hire from within so it might get you a job like it did self and bro.
>One way to get machine time is taking a CNC course at your local community college then being helpful (volunteering is fricking gold) in return for permission to use the machines. Schools like cool student projects and hire from within so it might get you a job like it did self and bro.
I used to do this, the uni I went to had makerspaces everywhere that you could just walk in and do stuff. The problem is I graduated and it's two hours away and so I'm stuck with doing what I can do on a desk. Working towards trying to rent some cheap shop but I'm learning as much as I can first before being stupid with money.
One can also do an impressive amount of work in an ordinary house or even dorm room with some thought put into storage. I had one and a half motorcycles in my military wall lockers and passed inspection.
>I had one and a half motorcycles in my military wall lockers and passed inspection.
impressive stuff anon
required classes? i told you my background is in CS (and in economics), not engineering. Tho I have done forward/inverse kinematics from game development same for pid loops when animating 3D rigs but there's must be more to getting something to move in the real world with actual materials or these robots wouldn't cost $75k.
>something more
if you take any robot/manufacturing/animatronics course youll know almost every part for Spot is likely custom built. it uses a jetson xavier but also likely on a custom board, as well as the motor drivers, bms etc. the motors are encoded linear in custom packages, all parts are designed to be replaceable and manufactured, they have braking systems in for safety stops, and not ragdoll when carrying 14kg. its has enough sw to do active 3d machine vision and conquer stairs, be autonomous as well as process the 12dof kinematics in real time. all that stuffed into a small user serviceable dog like chassis with a 90 minutes operating time, or 4hrs of idle. dude look at these specs and abilities of course its fricking expensive.
the Spot project if i had to guess, has probably north several million in house man hours, north of several million budget and is based on all the other work they developed for darpa and their other dogs for the last decade, with the help of hundreds of people, dozens of disciplines and thousands of 3rd party contractors. and it still took years.
how close do you think youd get emulating half of what BD achieved with Spot in a "makerspace" with a few smart guys, some 3d printers and a shoestring budget? any gorrilla Black person can make a dog toy with moving legs and a rasp pi for a brain. it took more work than a single natural life for BD to meet those specs and perform like Spot. if you took any pm courses youd realize with such meager means your scope would have to be reduced greatly which brings the question, is it even worth it? whats your minimum threshold off success here?
use that fat cs diploma of yours and draw up a scope. list resources and we can at least help determine if its even possible. dont be a dumbass and argue that a couple dudes and arduinos can achieve from scratch what took millions of dollars, hundreds of really smrt people and years to develop in a reasonable time
>is it even worth it? whats your minimum threshold off success here?
so just give up? what's the point you're getting at? technology improves with productivity yields so that you can leapfrog where it took millions before. OP obviously is never getting anywhere near the polish of a Spot robot, but he can definitely build a robot dog with an arm. You go back a over decade to when Boston Dynamics was working on BigDog
You can read the hundreds of patents that Boston Dynamic holds, for free on google patent, that didn't exist a decade ago to understand everything inside instead of working from a blank canvas guessing.
https://www.amazon.com/Unitree-Quadruped-Robotics-Adults-Embodied/dp/B07TZ3YC3N
you couldn't buy a quadruped like this for just $2k a decade ago but now you can, you can break this apart to speed up experimentation and what you need to learn. When the frick will you pessimist get off this board and realize you're not as smart as you think you are
https://bigthink.com/pessimists-archive/air-space-flight-impossible/
so basically this OP, you can just go out and buy a quadruped. It's not PrepHole you say? I'd say it'd be PrepHole if you really start hacking at the controller. Quadruped robots aren't that great honestly, there's a lot of room for improvement in their controllers. Getting faster running speed, longer battery life, and more terrain capability isn't out of the question even for an amateur. If you must build a quadruped or are interested in making shit that ain't quadrupeds you can buy replacement actuators for unitree quadrupeds on aliexpress. Here's a suspicious looking motor: https://www.aliexpress.us/item/3256806215791965.html
All these quadruped robot actuators look like they'd be pretty fun for diying robotic stuff that's not quadrupeds.
>op wanted to do it all from scratch in makerspaces with no budget and the power of friendship
>this Black person show scanned solutions
goddam this board is dumb
>>op wanted to do it all from scratch
where does it say OP wants to do it from "scratch." This is the diy board, not the reinvent the wheel board.
>how can I make a cake
>get cake mix and bake it at home
>okay thanks see you later
picture might make more sense if it was dated.
if this was a few years before the orville wright brothers took to the skies then sure good point, but if it was a few million years prior then it completely changes the point of the message from dont give up to just give up
it was in the article linked
>In 1903, New York Times predicted that airplanes would take 10 million years to develop. Only nine weeks later, the Wright Brothers achieved manned flight. The pathologically cynical always will find a reason to complain.
speaking of a few guys, arduinos and a 3d printer, this is what has been achieved with diy shit:
https://novaspotmicro.com/
look how much hes put into it already and how it pales in comparison to the real spot. its not even on the same fricking planet.
and then theres the 2700 dollar ripoff:
https://www.theverge.com/2021/6/10/22527413/tiny-robot-dog-unitree-robotics-go1
it can only carry a fricking water bottle.
and the alpha with basically the same specs amd roughly the same price.
https://www.indiatimes.com/trending/social-relevance/alphadog-chinese-have-copied-the-robot-dog-selling-it-as-ai-pet-537942.html
Then focus on getting your robot dog working in simulation first. Gazebo's pretty good. Mujoco and PhysX are open source now too.
kinematics
Isn't enough. You really need to be able to control force and more.
>>$75k
Actuators are the expensive part. Unless you have good actuators, your robot will suck. Boston dynamics can just mark stuff up because they can.
I learned to program myself in 2 years by hacking on microcontrollers and trying to write device drivers in C/C++. Programming isn't that hard to learn, especially if you have a specific, well-defined problem you're determined to solve, which you'd know from experience if you weren't a dilettante doomer homosexual
OP, if you have a CS background, you should be able to write the software yourself. Start with the Arduino servo tutorials, make it swing, then build from there. You're probably best off copying open source CAD designs first.
delusional
There's plenty of folks on YouTube who have singlehandedly built and written software for their own dogbots. Not as sophisticated as Spot, but fully functional.
Brother we wrote that shit in 3 weeks using V-rep during Robot Locomotion 102. Max plus synchronisation of legs you pick a gait pattern and it will walk fine. Basic SLAM and path finding is built into ROS or you roll your own in a few days with ultrasonic sensors
get with the times old man, you can let AI write the code. Also why are you even here?
Just copy this guy
good find anon, was looking for something like this. thanks
https://github.com/XRobots/openDog
https://www.instructables.com/Nova-Spot-Micro-a-Spot-Mini-Clone/
Build a minicheetah
https://build-its.blogspot.com/2019/12/the-mini-cheetah-robot.html?m=1
https://github.com/mit-biomimetics/Cheetah-Software
You can just buy the actuators off of AliExpress nowhttps://www.aliexpress.us/item/2251832799357101.html
>https://build-its.blogspot.com/2019/12/the-mini-cheetah-robot.html?m=1
this is incredible, good find anon thank you
Be ready to spend over a year or longer worths commitment
that's fine with me, I'm realistic with progress being slow
I was reading up more things today and found this project out that helps too
https://www.raspberrypi.com/news/freisa-the-plant-watering-robot-dog-magpimonday
I mean robo waifus are cool and all but the anons who aren't subtle about it are never really productive. I seen a bunch of dedicated robowaifu construction places and it's more cooler than diy doing. It's best to never mention anything about robowaifus but just have things that happen to be repurpose to turn into robowaifus
>cooler
coomer*
I have a discord but it's for a robot waifu and I'm selective on who gets in because I do plan to sell it. It's been quiet lately so it might have died. The discord not the waifu...
The collaboration part is the part that the op is misunderstanding. There won't be collaboration specially with channers. Maybe it it was German guy but probably not still. They don't understand the concept of meritocracy. You do what the guy who knows tells you. You losing your job as a web developer doesn't mean you have something to contribute. Neither is doing something stupid like trying to invent ai from scratch or sharing google links. If you want to learn ai you study stuff like pytorch or aloha ai or whatever Until then we're not equals.
>we're not equals.
crab bucket mentality, you said a lot of nothing. Let OP build his robot dog, yea it's going to be garbage but if he's serious he'll eventually get somewhere. it's miles better than rotting smoking weed playing video games
>if he's serious he'll eventually get somewhere. it's miles better than rotting smoking weed playing video games
this. never discourage a young inventor.
yeah because it might actually be a bomb next time
I still don't get why they made it so sexy
Find out where they get deployed in black neighborhoods. Then just go pick one up.
They're free, you can just take them.
Lookup: OpenDog
You have to learn how to move yourself and understand what you're doing before you'll get good results. Slow is smooth, and smooth is fast. Anything you can do fast you should be able to practice slow. If you can't bapance on one foot and know what you're doing to make it happen, you will be lost. This is why all of James Bruton's walking robots kinda fricking suck.
all you shit ass morons in this thread are all fricking stupid
click this fricking link https://github.com/reubenstr/zuko then click the other links in it and google the shit you dont understand
Neat I'mma frick with it. Got anything for simple shit like making a skeletal structure for a toy bird or something, getting its head to move back and forth and opening and closing its wings? I want to start with some simple toy things with preprogrammed motion that just runs and returns to default.
OP here. the github repo mentions
>SpotMicroAI
this is actually perfect, this is literally beyond what I could had ever expected. Literally lists everything to do step by step. Thank you so much anon. Also thanks everyone for all the links, going to read up on everything.