Asked
— Edited
Resolved by fxrtst!
Anyone building a robot dog? I am looking for Servo's to build one that will work with ARC...
Related Hardware EZ-B v4
Related Control
Servo Speed
Hey @fxrtst have you played with those 60kg servo yet? I see a lot on eBay for really cheap but I haven’t purchased one yet. $30 versus $300 for a dynamixal seems like a good deal. I wonder if you can do a potentiometer hack on these as well?
These are my new favorite servos..I purchased directly from AliExpress....and I can say is they are incredibly strong, one of the quietest servos I've ever used (comp Dynamixel)...also great hardware available for them too. My drink droid has this servo at the elbow and can lift the entire load of the forearm and hand. Because, of their success I chose them for upgrading my Hexapod, Gar. I couldn't see why you couldn't do the hack..as long as the pot is accessible. Then you have a cheap $30 servo with feedback.
Just be aware...these are 1/5th scale servos...so they are not the same size as a standard servo...must bigger.
@Will, your such a tease. Can you give me a proper name or a link for these wonder servos? I've seen you use them on your videos but if you pointed to where to get them I can't find it. So much info, so little time. LOL. I don't have the patents to wait till your next fab Feb video.
Here's a picture of the $350 servo I'm using now. (It's no longer available)
aaaahhhh I remember those Seiko servos..I think I even still have one I purchased in the 90s....been around a long time. Those Seiko are still available...I think you have the PS-050, they have the super massive one for $550! the PS-105 I think that's the one I have.
http://www.hobbyclub.com/index.php?main_page=index&cPath=24_225
The servos from my hexapod and the drink droid: 60 Kg/cm servo from Ali Express I'm pretty sure this servo is not strong enough to lift the bubble section...maybe two working in tandem?
Great thanks @will. I’ll give them a shot. I see you purchased metal horns. Why on earth do they sell a 60kg servo with a plastic horn. My next project I was thinking about a robot dog. I refuse to (and can’t afford to) pay $70,000 for a Boston Dynamics (now Hyundai) Spot. Pretty sure you could build a pretty good dog with those servos and the T265 for ~1000
Yeah I think so too. I’m not sure of the speed of the servos.. for the lower leg and forward stepping you would want snappy fast speed to get that leaving the ground steps that Spot does. Buy three servos and experiment with a single leg set up. Btw you can run these at 8.4v giving you a bit more speed and torque.
i have a dog on the drawing board...you’ve never seen anything like this before. I am experimenting with the CANBUS 3 phase motors. They are pricey but even with these Motors and controllers ($400 per axis) your looking at about $5000 for the project.. well below $70,000 and even some of the smaller dogs that are about $30,000 from other manufactures.
Moved this convo out of the intel realsense thread into here
The Lynxmotion ST1 servo's are pretty snappy but they can't carry much of a load. Nice build though.
Whhhhoooooaaaaaa, nice dawg!!!
Those are speedy indeed. Very small dog ( pup) heheh. I was thinking with some ingenuity You might be able to gear the output of the 60 kg/cm servos to move faster like a 5:1 . You could use a belt drive, chain or 3d printed gears. The lower leg could be strategically placed rod ( close to pivot point) to gain some speed.
For all you guys out there, looking for a new pet...
https://www.desmos.com/calculator/rpnk0wnpvj?lang=de
NEAT! Good find!
Are the servos you are referring to @Will the ones used on MIT robot dog. I see some on ALI for $500 to $900 each. You can buy all the parts or a complete Dog for ~6000 to 8000 USD if you search DOGOTIX MIC-01A (looks like there is an MIC-01B as well) but there seems to be a couple of companies selling them under various names / models. They are not smart (remote control) so would need some sensors and a brain. hmmm
I can actually get the MIT Mini motors for about $299 USD and then a Flipsky driver board for another $100...so about $400 per DOF x 12 +$4800. But those kits are intriguing....might be able to extract what I need for a build. Nice find.
Wow, the DOGOTIX MIC-01B is only $3250 [u]USD....hmmmmmm [/u] Edit..some places it says $3250 and others says $6000 to $7000 lol never a good sign! Maybe assembled vs unassembled...?????
Nink, Watching that Cheetah vid was interesting. Ont thing I have never seen in a walker happens at 2:26 when the knee joints invert and the robot swaps orientation from front to back. Seems so easy to get out of a situation rather than walking backwards until you are clear. That was cool.
I sent a couple of emails off. Wife is going to leave me for sure.
Those motor servos are amazing there is even a couple of bipeds being developed based on them.
Be nice if ARC had a driver for them.
This is a great time to be interested in Robotics.
@nink, My wife too...lol.
We can control these motors with a little third party help. They run on CANBUS, which is the system built into your car electronic system. The motors are 3 phase and they use Field Oriented Control (FOC). This creates a sequence for the coils in the motor they kinda fire off in sequence to get the shaft to turn. Because there is no gearing only electromagnetism holding a position there is compliancy in the shaft..that means it has a little bit of spring..like a shock absorber and keeps from tearing apart something like a dog that impacts with such force on the ground.
You can control these motors using many control boards available like the Flipskys that I used on the hoverboard hack on my channel. Those are similar to the robot dog motors. These controllers allow you to connect to the EZB or IOTINY or anything ARC supports using serial or you can use standard PWM like a standard servo control. Without an encoder they are only good for rotational movement, but with a hall effects encoder you can control positional. Then you have a massively strong servo to build a robot with.
$300 for motor $100 per flipsky or other FOC controller , so about $400 per DOF.
Here is an example with a very cheap ($20) FOC controller and an Ardurino with a sketch controller....this acts as positional motor/servo. EDIT: Seems this board is no longer available..and limited by the Atmel onboard.
@fxrtst that is a VERY good find!! I am still having this encoder and and few brushless motors, so I might actually build one of these! That is a nice Christmas present!!
I contacted Dogotix and the only place you can buy is through Alibaba unless you want a large customized order. There are two dogs on there MIC-01A and MIC-01B
MIC-01A $6500 The MIC-01A has stronger motors and can last 1.5 hours, has the ability to add a realsense D435 (not included) and a Jetson NANO TX2 (not included). Neither the D435 or Jetson Nano are supported today by ARC. There is a sample? for the MIC-01A for $4545 US but actual $6596 CAD (Robot+ Transaction Fee + Credit card Fee) +13% Tax =$7453 CAD https://www.alibaba.com/product-detail/Medium-Electric-Driven-Robot-Dog-Based_62491310900.html?spm=a2700.galleryofferlist.normal_offer.d_title.12e06784Vbvz5z
MIC-01B $3250 The MIC-01B has smaller motors so can handle less payload and walks slower, It runs linux and MIT opensource software but it does have an intel ATOM PC but not sure how much RAM. I guess you can hack it and add your own sensors and get it to work with ARC but that is a lot of money to drop down if it doesn't work. The MIC-01B for $3250 US but actual is $4872 CAD (Robot+ Transaction Fee + Credit card Fee) +13% tax = $5505. (Ontario sales tax). https://www.alibaba.com/product-detail/Medium-Electric-Driven-Robot-Dog-Based_1600144281544.html?spm=a2700.galleryofferlist.normal_offer.d_title.e8646784EXWv1u
Takes 3 weeks for delivery.
@DJ you only pay 5% sales tax :-)
If anyone buys one please post
They don't have stock ???
Messaged seller and got this response. Looks like they are building a new range.
@Nink:
You need to be very careful with internet shopping. I ordered a German shepherd and now this man is living with us.:
@PTP, LOL. that's hilarious!
hahahahhahahahhahahahahahh!
Well these new dogs can only be better than their predecessors...probably worth the 1/2 year wait?!
Looks like this guy built his own version of the MIT robot. https://jpieper.com/2020/11/09/mjbots-november-2020-update/ If you watch his November update video and follow along with his blog you can see his journey, he developed his own motor driver controller board and motors as well.
MIT Thesis here https://dspace.mit.edu/handle/1721.1/118671
This time around I ordered (I Hope) one of the developer kits from MJ Bots Figured this would be a good place to start. supports CAN-FD and can run up to 400W. About $200 US including shipping. Actual Boards are only $84 if someone else wants to get one but I figured I needed a motor PSU connectors etc and saves me running around finding parts. https://mjbots.com
Interesting. The moteus r4.5 controller is quite a bit cheaper than my flipskys FOC controller. Let me know how much torque you get with those motors. I was planning on something quite a bit bigger.
Josh is amazing and has really come a long way and is an excellent engineer and developer.
There is a good video on the moteus dev kit config. USB connect to PC and Python app controls.
Is this the same dog you all talking about? I just read this, they say now anyone with a printer ,500 dollars in parts and something like Raspberry Pi,can build it "Chop"... Dog,Spot
Nice find. Looks like a different dog the original one I posted used off the shelf lynxmothion brackets and servo's but 3D printing could be a lot cheaper. The 20kg PWM servo's are cheap but kind of slow , a guy could die of thirst before the beer arrived.
The moteus r4.5 controller looks good and seems to be easy to setup...this clip above is just some nice explanatory stuff on FOC!!
Guys, can someone tell me the advantages of using this kind of motors over stepper motors?
These motors are faster and more powerful...
There is a patent app for the FOC stator motor with internal gearing that was used in the Robot from MIT, it hasn't been granted yet but looks pretty thorough so I think we will see a lot of these startup robot dogs using this actuator shut down although I am not sure China has a lot of respect for our patent laws. https://patents.google.com/patent/US20200230811A1
Looks like there is more stuff on the way... 2021 will be full of surprises I guess!!
Good find..there are FOC controllers coming out everywhere...! Everyone using ESP modules for wireless...these days.
FOC shield and code for Ardurino: https://docs.simplefoc.com/position_control_example
Bu this board is limited to 24v max
Been scouring the internet for over a month on this FOC subject with my eye on the prize...position control with PWM. I've joined quite a few discord servers to discuss with robot builders and folks designing their own controllers or expanding upon the SimpleFOC from above and what I've found there is a hard line division between pro hobbiests -those who want to play around and experiment with standard RC controllers and FOC motors, and pro robot folks that control FOC motors via ROS or other means.
So to get FOC motors to work like a servo and with skills in ARC, it would be nice to have a controller that can control precision position not just velocity. I found a card called the NearZero. Its home brew but open source. They don't have any cards available at the moment..as they are moving onto ver 2. Only draw back with current board is a fairly low max current at 3 amps per channel (2 ch per board). I've shot off an email to see what the specs are on V2 (due in Feb).
Start at 11:30
My dev kit turned up this week happy with initial impressions. I try to have a play on the weekend.
....be sure to join his discord....very active and he seems to be online and available most of the time...
Do you have discord invite links?
..sorry late reply..i can see you in the server now
This is another one to check...
@Nink which dev kit were you getting?
@all I ordered two of those Simple FOVs now, and will compare them to my ODrive...it seems like the Simple FOV is designed with small BLDCs in mind, but that would no worry me too much! Could be nice for anything not needing a massive amount of torque!!
@Mickey666: post #30 https://mjbots.com/products/moteus-r4-5-developer-kit
I'm playing with dev kit and so far is working out of the box on Linux and Raspberry PI, the overall design is very good, the controller interface is CAN FD, single motor and the pcb has an absolute magnetic encoder so the idea is to pair a controller with a magnet attached to a BLDC motor.
The dev kit has everything included to start: 24v Power supply, USB-CANFD, Controller attached to BLDC motor.
The solution allows to control a BLDC motor like a 360 multi turn servo, plus all the other stuff: torque, velocity.
I'll create a plugin to interact with controller.
@Nick: Thank you for introducing it to me, did you manage to test it ?
Hey guys...since this is getting discussed again, I thought I might share my super chopped up demo code for ARC and Odrive!! This is just a POC to show how a BLDC motor on closed loop can be controlled with ARC, the firmware basically sends all servo input from ARC straight the M0 on the Odrive, it is more or less just a starting point and meant to be developed further! I thought it might come in handy...
https://github.com/SwannSchilling/ODrive-for-ARC/blob/main/ODriveArcArduinoTest.ino
@Mickey666 Nice!
You can use the AS5047P absolute encoder: https://www.digikey.com/en/products/detail/ams/AS5047P-TS_EK_AB/5452344
with odrive. https://docs.odriverobotics.com/encoders
Using an evaluation board simplifies the integration.
The AS5047P is also used by the moteus controller.
@mickey, I saw your post on another page about the torque you are getting with your set up. Have you noticed any heat in your gear box from the motor?
Great job on the ODrive set up....you should make a video on your set up!
@PTP The dev kit worked fine in ubuntu but I couldn't get it to work properly in windows, I posted in the discord. I saw above you are looking at making a plugin NICE that would be very helpful (You are amazing).
@Mickey666 Nice work and demo on the Odrive working with Arduino and ARC. Very cool.
The reason I went for the Moteus was he had made the opensource dog and I figured that would be the easiest way to get one working with ARC ( I still am not able to buy a dog, so only choice is build at this stage unless someone wants to give me 75K for Boston Dynamics Spot) . I like the nearzero idea as it uses standard PWM or I2C so it should just work out of the box with ARC.
@nink I'm following on the discord. Let me know if you figure it out. I never did hear from the developer of Near Zero..i didn't think i would. Some flakery on the home brew of this project.
@fxrtst I guess what we would have to to with 3d printed gearboxes is to stresstest, I did not do this yet. All options have pros and cons! I went for the Cycloidal because of the low/zero backlash, but it is taking some effort and money to build with all those bearings and other parts! Planetary Gearboxes on the other hand can be easy to print, but they might have some backlash and I don't know how good those teeth will hold up, when exposed to a lot of torque. The choice of the right gearbox is as important as the motor/driver solution, it might be strongly dependent on the scenario of your application! If there is a cheap option available that is assembled and ready to use, please share!!
The Moteus is a very good find, I am more than tempted to give it a try. Both the Moteus and the ODrive have pros and cons... The Moteus is very specific for being used on a robotic joint and to be used straight out of the box...eg encoder is on board! The Odrive has a bigger form factor, but lets you choose the encoder and is more a general purpose dev board!
Odrive setup for ARC is pretty straight forward, I dont own a Moteus (yet?)...so lets see where this takes us!!
..yeah its the wild wild west of controllers for BLDC/FOC motors! Everyone is making them and alot of them are useful for only certain sized motors (except ODrive).
This guy has a very easy to understand series on YouTube and shows different use cases and set up for open and closed loop examples... yup you guessed it with HIS version of a controller he developed...its got two channels for $45USD but again limited by max current (about 6 amps with 30A short bursts). Also has a built in Arduino. So definitely depends on the size of the project you are building. Watch his whole series..!!! Good info.
@mickey are you on Discord?
I heard back from Justine (who is an amazing person BTW) and all about ver 2 of her board NearZero2. It reads in part:
Sure... hit me up
Mickey666Maus#3529
Btw I posted a video of exactly that guy #45
It is more or less the same like the SimpleFOCShield, both are using the L6234 IC and just come in slightly different form factors! The SimpleFOCShield is, as its name says, a shield to be used on an Arduino or compatible board... Random Access Projects is having the it all on one board which might be nice since it is more compact!
So again, its the same choice, one is more a def board since it allows Arduino, STM32, ESP32 and Teensy boards to be setup. The other has a pretty nice overall package but you are set to what it is... It might boil down to your application or maybe hardware preferences!
Btw, if you are interested in driving bigger BLDC motors with the SimpleFOClibrary, it is also quiet possible...
...it seems to be raining BLDC/FOC controllers. Lol. Yeah I think that it is dependent on your applications. I look forward to PTP plug in solution for the moteus. BTW Skyntific looks to be testing the dev kit....hes been active on there (discord), trying to sort through some issues he has.
I def interested trying out the Near Zero2 when it comes out.
I must say that I don't know too much about the differences in hardware when it comes to the Moteus and the Odrive... Both seem to offer very good documentation so its pretty easy to figure out what to send to those registers if you want to write up some code! Both seem to have a vivid Discord community, so these are pretty good options!
The other boards are very tempting because they offer a low cost option where you don't need a ton of torque, but some of those advanced control options and high speed might come in handy!!
Super interesting, I am happy this is still hot gossip!!
P.S. if your are wanting to try the less complicated approach of a Planetary Gearbox, this is a good starting point...
http://www.thecatalystis.com/gears/
Plus the plugin for Fusion 360
https://apps.autodesk.com/FUSION/en/Detail/Index?id=9029586664984391977&appLang=en&os=Win64
Or give this super simple Cycloid Gearbox a try, it is also very basic like the above...but both examples might offer a good entry point to 3D printed gearboxes!!
https://www.thingiverse.com/thing:4604783
Great addition to Fusion......looks like something I could get lost in for days! (gear maker)
I'm already in the early design phase for my robot dog...what sensors do you see a robot dog having and for what purpose would they serve?
I think sensors are really going to depend on use case for the dog. Example if it is used in mining after a cave in to check for survivors, you will want a photo sensor (are the lights still working if not provide light), Oxygen and Carbon Monoxide sensor (is the air breathable), vibration sensor (Stop wait check for movement), Directional Microphones (listen for sound), radio tracker (Is there bluetooth, cell or other signals), Lidar to check height and width (Will the dog fit or be able to walk in that space), location track like T265 (what route did I take, how do I get back, how do others follow). I am sure you could think of others.
I think the best way would be to have some form of sensor platform and you can mount on the sensors required for the use case you need.
A great video this week by James Bruton. He has sussed out quite a bit with our T265s. This is great for robot dogs but this video is applicable to DJs navigation system too.
Thanks for sharing Will, I try to follow along with all of Bruton's builds although I must admit I was a little confused why he put his T265 above the LIDAR obstructing view. I am struggling at the moment how to mount my robot arm above my Lidar as I need strong support base but can't obstruct LIDAR view. I think I will just have to put LIDAR at one end and Arm at the other so it minimizes interference.
Back to the FOC here is a good MATLAB on how FOC works with BLDC and PMSM motors. (You can go back to episode 1 to 3 for BLDC PMSM differences, PWM, Brushed motors etc)
Thanks for the above, although it is a bit hard to comprehend all the information, it is really good to get familiar with the terminologies!! Good find!!
Setup guide for the ODrive and ARC
https://synthiam.com/Community/Robots/BLDC-Control-ODrive-for-ARC-20199
Awesome! Looking forward to trying this out! Finally a use for my ODrive!!
Let me know if you are running into any isses!!
So this one seems to have a decent pricetag, while maintaning all the performance of the high priced counterpart...
https://tinymovr.com/collections/all
Nice find. Very small, ideal for animatronics where you need small whisper quiet servos and components. I have been looking at some 22mm BLDC motors will HALL sensors this would be ideal for.
Awesome...please share, I might get some smaller sized BLDCs too!
I am working on a joint for the use with those Tarot 4108 BLDCs, got it working and will post once I can tell more about the progress...
Usefull stuff!!
https://apps.autodesk.com/en/Publisher/PublisherHomepage?ID=HNK2PN6ZY74X
Look Ma no sensor...lol. Someone also added to the firmware to get around the 360 limit on VESC so you can now do a multiturn for things like gearboxes, called MJ-VESC.
Nice, I did find someone also building a VESC controller over at Hackster...but I am not sure about how good this is for actually getting accurate positions! I mean we are in the robotics department and not up to be driving a remote controlled car, right? But I guess @ptp will know...I am really bad with this kind of hardware evaluation stuff!!
Btw...I just finished the assembly video of my gearbox, and will put all the files up somewhere, just in case anyone wants to follow along!!
Uploaded all the files for printing, also included a parts list...
https://www.prusaprinters.org/prints/53681-cycloidal-drive-3d-printed-gearbox-401
First I'd like to say that I spent my life growing up in my grandfather's motor rewind shop, which was our family business (and also a commercial wiring company). So I spent a lot of time around motors. Some small enough to fit in my hand, to motors so big, I could walk inside the stator at 9 years old! (motors 1.5 meters in diameter! ) I would play with all the steel ball bearings and mess around with reels of copper wiring. So all this new tech of FOC motors and windings, magnets and phases is exciting as it takes me back to those old days.
I thought I'd add what I've learned researching these BLDC FOC motors and controllers.
Most of the BLDC FOC motors we are talking about, are built for drones. Their main purpose is to spin props at high speeds.
The controllers we've talked about here, are to convert these motors into low velocity and closed-loop positional motors so we can use them like servos.
These motors are rated in KV. KV refers to the constant velocity of a motor. It is measured by the number of revolutions per minute (rpm) that a motor turns when 1V (one volt) is applied with no load attached to that motor.
4) Torque is determined by the number of windings on the armature and the strength of the magnets. A low Kv motor has more windings of thinner wire. It will carry more volts at fewer amps, producing higher torque. A high Kv motor has fewer windings of thicker wire that carry more amps at fewer volts.
We will most likely be using these motors as actuators, so look at a lower KV value (100 KV or below). This will give more torque to move heavy objects/robot parts. If you go with a gearbox, you can pick a higher KV rating, which has a smaller physical size compared to the larger lower KV motors and can often deliver more torque. But you gain backlash and noise in the gearboxes.
Cogging. This is something to avoid and is due to fewer poles and magnets in a motor. The result is chunky moves at low speeds and the inability to makes small incremental degree movements. But, some developers are writing software for controllers to avoid cogging on these lower pole motors! A fascinating workaround.
Anyways hopefully this is helpful.
Used this ancient toothpuller to move my motor shaft...so now I will be able to mount the encoder on the other side of my BLDC!!
I will make a branch for this design, with the encoder to the back...and upload the files as soon as I am happy with them! There will also be a mount for the AS5047P-TS_EK_AB...
Tooth puller hahaha. That I know from my childhood is a bearing puller.
Everything is looking good. I look forward to seeing all axis work together. Is there much noise when they work together?
Actually everything is pretty good, considering the noise and the vibration... I will try and make a torque, speed and precision test asap!!
But today I was having to deal with a clogged nozzle, you know the drill!! Man am I happy, I did not lose some screws on the way...
Haahahha lots of fun....!!!
Just had to throw this one in here...its so well filmed, sometimes you'd almost think it is CGI!!
Wow. I want one. Just think owning a dog like this and not needing a pooper scooper.
I want one as well, I just lack the $70,000 needed to buy it.
@Nink, sell the house to pay for it. You'd be one of the coolest homeless guys around. LOL.
If I didn’t have a wife and kids... Maybe we should just short GameStop. You go first.
Adding battery info, because if we order motors to test with, we need them to run!
If your project is large like a dog..make sure you pay attention to the voltage requirements for that motor because you will get better performance with max volts. As an example...drone motors are using standard hobby lipo batteries..but depending on the size of the motor voltage can go way up. Try to match the battery for your motors.
Standard lipo battery voltage for drone motors: 3S is 11V 6S is 22V 12S is 44.4 V 14S is 48V
How long your robot runs will depend on how many amps you can feed it. Remember If you have a 1 amp/hour battery (1000mil/amps) you can run a 1 amp motor (no load) for about an hour. If you put that motor under load and it pulls 2 amps, the battery would only last for 1/2 hour and so on. On a dog if you had 12 motors, pulling 1 amp at idle, that's 12 amps/hr. So if you had a 12 amp/hr battery you could run it for an hour. But we know the dog will be jumping running and the current draw will be different every time you take it out. Just get the max Amp battery you can afford...they get pricey the larger you go!
You can also get 2 batteries and combine them in parallel or in series for more volts/or more amps whichever you need.
Something to consider is a voltage step down to get more amps out of a high voltage battery. It's usually easier to find a high voltage battery than high amp batter. If you need 12v but use a 24v battery, you'll double the amperage and runtime. But be sure to find a digital switching power supply that can do it.
Here's a 10amp 24 to 12 one: https://www.daygreen.com/collections/24v-to-12v-13-8v/products/24v-to-12v-10a-120w-dc-dc-step-down-converter-voltage-regulator
And 30 amp: https://www.daygreen.com/collections/24v-to-12v-13-8v/products/24v-to-13-8v-30a-414w-dc-dc-step-down-converter-voltage-regulator-1
What? 30A ?That is awesome, I was always putting 3 or 4 12 volt batteries together in Parallel to get higher Amps, longer run time on large wheeled bot,what a wasting of batteries and trying to charge them all, I should have just asked you last year how to do this easier!
Just remember...
It is going to be fun when we start wrestling with robots running several 24v 5amp BLDC motors and even some 48v 10amp motors. Thats a lot of 2S Ezrobot Lipo's.
You can now change the ratio of my gearbox on the fly...
And some motor stuff which comes in handy at times...
https://build-its-inprogress.blogspot.com/2016/01/motor-characterization-for-small.html
Great Dope Tech Video. Looks like they have a choreography service. https://github.com/boston-dynamics/spot-sdk/blob/master/docs/concepts/choreography/choreography_service.md
Yeah man, sometimes I am still thinking this must be CGI!!!
There's a new Adam Savage video that shows the new animation timeline tool for Spot. Motions are blended together to morph from one animation to the other. The timeline is nearly identical to the one I proposed in 2015. Like a nonlinear editor. I've always said robots would become more acceptable by how they move and how they look. A keyframe/IK/motion capture combination is the key. Motion blending is the second key. Make robots move naturally and people will go nuts for them. 27 million views on the BD robots dancing says so.
Absolutely amazing!!!
FYI new extended feature launch tomorrow. https://www.youtube.com/watch?v=WvTdNwyADZc&feature=youtu.be
I have been looking at the API and the ROS integration with Spot and I think a ARC plugin could be written fairly easily to provide some basic functionality. Also SPOT has a sensor platform where you can add OEM sensors (Lidar, Robot Arms etc) so ARC would provide a lot of value.
So who has $75,000.
Adding a few more bearings and now also a second encoder per joint... I need to do a testing rig, but it is looking pretty good so far!
These guys managed to get very precise with their Cyclodials...
...now all we have to do is line up for a Markforge...heheheh (Onyx 1 is $4500 and the Onyx 2 is $10,000) Wonder how much a roll of carbon fiber filament costs?
Its a bit more expensive than ordinary filament and you need a hardened steel nozzle...there is no magic to it, any ordinary printer can print this! Its just a bit pricey for prototyping that's all!
How about another controller...this one called the Dizzy..controls BLDC motors like servos with 180 degrees.....
http://www.dizzy.ai/
That is a VERY good find!!! The controller is only 35$ and it looks pretty amazing...with the encoder build in an hollow shaft being supported by design!! I am following the guy on YouTube but never realized, he is also selling products!!
Great...this could help to reduce the per joint cost! Quads are getting exponentially expensive...just way to much joints!!
Nice find. This would certainly bring the cost down. Anyone know the compatible BLDC motors. His Robot design doesn’t use a drive belt like the other designs (i guess you could change this) and all wiring internal. I do worry the continual twisting of the wires would cause them to break over time especially when you have 3 motors back to back (maybe that is why he put the 3rd motor in the middle of robot leg).
@Nink, Should be compatible with any motors in the controller's voltage range (10v-40V). Belts and gear boxes are giving you all the extra torque...you can run right off the motor (lower KV) if your robot is small enough without all the gearing. Also as long as you have a slip ring like he demoed wires should be ok...especially because his controller is set up like a servo at 1-180 degrees vs multi turn.
Edit: Also looks like he is sold out of everything....and not been any for awhile...he has posted everything open source so you can build your own board....sigh
The video shows you need a BLDC motor with a central magnetic encoder at core. Some BLDC seem to have hall magnets on the circumference not in the center and others seem to have no magnets for hall sensors at all.
Yeah I was thinking you could just use the special magnet like the one included in the MJBots dev kit and it would work with his magnetic encoder. But with out the hole in the center, you would have to keep your wires outside the motor, which won't be a problem. I think maybe its because he is showing it with the controller rotating with the motor at calibration at 2:37...which it doesn't when installed in the leg..see the video at 6:00 mark. Now the motor is inverted and the part of the motor with the controller wires is on the bottom and not moving with the motor. He has wires going through, but you could easily have them outside of the motor with no problems getting twisted up.
Hey, I also noticed that the guy is no longer active...I guess that is also one thing that we should keep in consideration. If we are going for a cheap diy version, there is always the possibility of grinding to a dead stop when it comes to development. I would not mind to get those PCBs made somewhere, I mean you'd need a few anyways. So the price might be OK?
But if there are any issues, there is no one to give a helping hand and you'd be there by yourself...
So I guess, unless there is someone with enough knowledge popping into this adventure, it would be not the very best option!
But man, the ODrive and the Moteus will cost a lot of money if you are going to build a quad! For a 4DOF arm it is fine, but a quad will be much much more!!
Btw...I moved the encoder to get a bit more room!!
Trying a new BLDC style joint...lets see if this works!
Outdoor stair climbing confrontation: https://twitter.com/HomeiMiyashita/status/1335856042509365250 Unitree A1 vs Spot
If you get a prototype working - I’d recommend reaching out to ezrobot and see if they’d take on manufacturing. Last I spoke with them, they’re starting to focus on diy market a bit
I guess the one off the self D435 on the A1 doesn’t compete with the 5 stereo cameras on BD
what amazes me is you can buy a cheap TOF laser sensor for $10 so putting 1 on each leg isn’t exactly going to break the budget of a $10K dog.
I'm guessing the Spot's front maps a wide area with much more detail (2 depth sensors).
Spot can climb down using a single 3d sensor so I believe the difference is their software and expertise. You can cut corners and costs but in the end the software is the key.
I am watching this project at the moment as it uses the Moteus controller but uses cheap 8308 BLDC's That puts the project in more of the grasp of the general consumer. If we don't use the expensive and heavy motors and go with 12 8308 ($67 on Aliexpress = 804), 12 moteus ($84 =1008) FOC, a canbus usb ($110) a PDU ($150)and use a lightweight X86 computer ($100) we should be able to come in around 3K with a couple of sensors. https://hackaday.io/project/176726-stanley-the-capstan-based-quadruped
Damian Lickindorf is awesome...I really dig his robots!!
Robin Fjord Dog is a MIT clone and this is very cool. Also uses Moteus controller and AK10-9 Motors but at $700 each ummm https://hackaday.io/project/176487-k3lso-quadruped
Yap...pretty awesome but also super high end!!
I got my motor in about two weeks even with the Chinese holiday. This is the eagle 8318 motor a lot of people are using for midsized and smaller dogs. Boy its quite heavy. Twelve of these will add a lot of weight to a dog. I'll do some testing with MJbots board this weekend. I am leaning toward a pully and belt to increase the torque and the ratio vs gearboxes.
Chinese holidays are over, as we can see in this picture too... Keep us updated on your progress!
I guess for a quad you will end up with a belt and pulley plus gearbox combo...at least that's what Josh is using for his dog! I really like those planetary reducers he is machining!
The motors I ordered are for a gripper, lets see if I can get it to work too!
I've been following this open source effort as well.
They have a website here. https://github.com/open-dynamic-robot-initiative/open_robot_actuator_hardware/blob/master/mechanics/actuator_module_v1/README.md#brushless-actuator-module-v1
Here is their main page https://github.com/open-dynamic-robot-initiative/open_robot_actuator_hardware
This is nearly identical to my designs. Double gearing should give a good ratio for speed and torque. I’m using skate board pulleys and belts. They are standard from the bearings to shaft sizes.
I like the idea of 3D printed with skateboard bearings and low cost BLDC. I wonder about the custom timing belts, I have to imagine they could get pricey. There must be some common off the shelf timing belts that could be used.
I guess the robot arm would need 3 of these actuators, the base if mounted horizontally won’t have a lot of vertical torque so could base just use an ungeared BLDC? The wrist and gripper will only have to lift the weight of the load so I guess I could use common hobby servos.
I have decided to base my robot on the TJBOTS quad build as there is a strong community and lots of support, although the price for parts is steep, especially if you use Josh expensive motors. My logic was to use the supported JSON to communicate with the quad so I don’t have to fork the base design and use a payload adding an SBC to handle all the sensors, additional actuators as well as command and control with ARC.
I will need to make a robot arm and gripper for it, but I don’t think the dog can carry a heavy load so what ever I do needs to be light weight and easy to attach / remove. Hopefully I can use USBCANBUS to drive the robot arm with the added SBC and ARC
@nink Here is one place I've been shopping. Very affordable. Here are kits and here for different sized timing belts up to 430 mm's. A google search will pull up a ton of belts and pulleys in regards to powered skateboards. Way more affordable than a gearboxes, but you need more space for this kind of a setup.
Very nice!! Good finds guys!! Looking forward to see where this is heading!!
I am about to finish the design for my end actuator...it will be driven by 3 small Gimbal BLDCs, hope I can get it all to work soon!
Will throw it to the printer now!!
Love this design...so slim!
I have been following Paul Gould for a long time now!! He is my hero when it comes to 3d printed gearboxes!!
Same here, Paul was the first person that got me interested in cycloidals way back when with his early walker work. He publishes his design files as well which is great. He was also responsive when I emailed him with questions.
A lot of torque with a 40:1 gearbox, but seems a little slow due to that. You would have to have an incredibly high KV to make up for the loss of speed in the gearbox. As we know these robot dogs legs have to be very fast and snappy to get that zero G balancing moment in their gait. I’ll be interested in seeing how this slim version compares to this one.
My gearbox is doing pretty good, its a shame I wrecked that motor, would be nice to get going on more DOFs and to show some tests on speed, torque and precision... But my setup is meant to be for a stationary robot arm and not for a quad! I guess for a quad you would have to go for a combo of planetary and belt or cyclodial and belt! Paul has done some nice Capstan Timing Belts lately!
This one looks great btw...but I dont know if I can do the math to get it right!!
This is a genius design.
If you read down in the description there is an expired patent on a similar design with only three points like a planetary alignment.
[url=https://patents.google.com/patent/US5954611]https://patents.google.com/patent/US5954611
Turns out I haven't commented on this post yet. I'm intrigued and I'm commenting to be updated on the conversation
...welcome to the convo..lots of testing and exploring..!!!
We have been waiting for you to jump in @Jeremie with a low cost FOC for the BLDC's and gearboxes.
Btw, Tinymovr is also about to release its version of an open source leg for a robotic dog...we had him in our BLDC controller checklist, I am on his Discord and he seems to be actively developing. So it might be something to keep an eye on!
https://tinymovr.com/
Also I thought, I'll just share this...just in case you guys wanna laser eyes the list!!
https://github.com/cajt/list_of_robot_electronics
@ptp that patent looks really cool, maybe I will try to figure out how this works...but to get the belt length right might be super complicated! Could be fun though!!
Just had to put this here...its too sweet!!
LOL, I love these old instructional toons. What do you think? From the late 50's early 60's?
New Printer Turned up, I remember last time I assembled a printer I said never again and yet here we are. Instructions said 8 hour and yet 10 hours later .... now I need to do the wiring but after half a dozen beers, I think I will walk away until tomorrow.
Nice one @Nink congrats to your Prusa!!! That one is a beast...never failed me since I build it, and its been three years so far!!
I built my first 3D printer (a Chinese Prusa clone) from a kit. The good thing is that you will always know where to look to fix it in the future. It is an invaluable experience.
@Mickey666Maus, is that you and your video? I know this is off subject but what video editor are you using and is it hard to get the fast forward motion? I've made quite a few videos and have struggled getting that effect. I'm using VSDC Video editor.
I really want to get into 3D printing and probably pop for a nicer one I'd have to put together for the reasons Perry_S mentioned. However fear of constant maintenance, adjustments and the learning curve scares me.
Nice Video thanks @Mickey666Maus. For years I thought I just didn't have the patience or skill for 3D printing and now I am actually angry. For years my definition of calibrating involved clamping on a hunk of glass with 4 bull dog clips and then adjusting 4 wing nuts using a business card jammed under the print head in a dozen different locations and you Prusa folks JUST HIT CALIBRATE AGGGHHHHH.
First print
Auto bed leveling was a game changer for me.
@DaveSchulpius I am using DaVinci Resolve 17, its free and one of the best Video Editing tools I ever touched! Don't get scared by the learning curve. Once you started printing, you will get hooked anyways... and maintenance is really marginal!!
https://www.blackmagicdesign.com/de/products/davinciresolve/
@Nink OMG that fresh printbed makes me instantly fall in love!!! love
Thanks @Mickey666Maus. I'll give it a try.
Resolve is all I use now...best editor.
Wow, I downloaded and Tried DaVinci Resolve 17. Wow, it's great. I picked up on how to use it right away (with the help of a youtube how to video).
Thanks!!!!!!!!
are you guys using a paid or free version of fusion 360?
Free!! Just renewed my free plan...best software ever!! Except for ARC of course!!
While building my little inmoov I have been working with gearboxes a lot. This isn't really appropriate for a walker actuator but I wanted to share a technique I am working with. Now that I have a SLA resin printer I have been printing the gears with that. The resolution is crazy so you do not get any of the layer lines that come with my regular printer and they mesh silky smooth. These early tries just used simple cosmetic resin and have lasted quite a while through testing although they are starting to show some wear (whiter areas). I have since ordered some high strength engineering resin to reprint these. We'll see how it goes. The rest of the gearbox is PLA, Bearings added to fix some of the original design wear zones. The worm support is really bad out of the box.
Nice worm gear. I am curious to hear about the high strength engineering resin works out. I used to use ABS and acetone to smooth out prints but you lose a lot of detail and it is not very strong. The PETG seems to have a fairly good finish although I am not sure I am up for dichloromethane smoothing.
Nice...good to know that resin works for stuff like that!!
@perry, all my resins are brittle out of curing box. Are you using something like TUFF mixed in with the regular resin to toughen the cured print? How do you tap it for the screw that holds onto the servo? I couldn't imaging using it for functional parts.
@fxrtst (man I'm still trying to figure out how to pronounce that and what it means)
We all know resin printing from the awesome miniatures with incredible high detail. They were $2000 one Christmas and $220 the next. They were almost all made for the miniatures markets. Some time has passed and there is a growing movement to develop higher strength resins and processes. You are completely right about the brittleness in parts printed with regular resin. Mine are too, compounded by the fact I am usually printing thin parts at the time. But these thick heave parts printed non-hollow have held up well. I am noticing they do not have the same lubricity from the material I get with PLA gears so we will see how this new resin will perform. It is a ceramic no-fire resin with high surface hardness which I believe will help with galling and a much higher modulus of elasticity which will yield a better stress/strain profile. As I said, all words until I run it rough for a while.
Resins are coming along. Here's an example
Follow up post....I just got it.... FX Artist
I'm so slow some time
wow they are approaching PETG. I ignored the resin market as I thought it was for creating shiny bobbles but @Jeramie brought it up last week in his car mod. So I was looking at the Anycubic photon model X as a medium size resin printer but was put off by the low tensile strength. Maybe time to reconsider for quick high resolution prints for smaller parts.
Reading https://www.3dhubs.com/knowledge-base/sla-3d-printing-materials-compared/#standard
Nice link Nink. Man I can can make flexible tires!
I got a Mars Pro for Christmas. Really happy with it and a lot of online support.
I have a review video on my channel for the anycubic mono X. It’s a great printer. But I’m getting very interested in some of these new resins. Most standard resins are urethane. I’m slowly stocking up on various mix ins like Tuff which adds a small flex to the parts making them less brittle.
Yes FX Artist. You got it!
I just repaired my damaged BLDC, so I thought the clip fits perfect into the thread...
Nice: if I didn’t know better I would think you are learning in preparation to design your own BLDC.
I actually learned a lot by doing this...and I am a lot less afraid now in case I will have to do it again!
Its amazing how simple and elegant the whole concept of BLDCs is!! love
As I stated earlier in this thread I grew up in my grandfathers motor rewind shop and every motor on the planet works the same way. Some differences: single phase and three phase windings, ac and dc. We had a cool machine that would create the windings then each coil was placed by hand in the stator, then the stator was dipped in resin and baked. The resin vat was built into the floor and held 100 gallons of resin. It had a steal door on top but I always envisioned as a kid falling in it and being preserved like mammoths in the tar pits! Lol.
Anyways, You did an awesome job rewinding it.
Awesome 3 in hand rewiring. We have a resin tank in the factory just like Will's dad but it holds 10,000 gallons!
Here's a pic from my day job, installing a 2.1MW BLDC used as a generator on a wind turbine. The team I work on did the design and prototype testing.
Awwww I love that! The biggest motor I remember, I was about 10 years old we had a stator that was big enough for me stand up in arms out stretched...but this one is waaaaayyyyy bigger. I've never seen a wind turbine apart like that before..so neat! Thanks for sharing!
OMG that is some huge a** generator!!! You always forget industry builds when working on the small scale hobby projects...amazing!!!
anyone working with PolyCarbonate. I am printing the MJBot components and the lower legs are PC Blend. Reading the docs on PC Blend looks like you need a hood (I guess I can make a makeshift one using clear plastic and dowel) but wondering what peoples experiences are. Also looks like the PC Blend is quite strong (although prone to warp on large prints) wondering if suitable for gears.
@nink Just be sure your hot end can get up to 250-300 degrees. It warps and splits like crazy..more so than ABS. Need to put the printer into an enclosure to keep it heated to reduce warping. Also PC likes a bed at 145-150. Print slow for first layer like 20mm/sec. No fans on.
Edit: I missed you were using a blend.....these suggestions are for PC by itself. Not used a blend..what is the other component?
Thanks for the info @Will. It is Prusa own PC Blend filament. I haven't ordered it yet but wanted to understand how complicated this really was. Sounds like the Prusa printer supports it so I am game, I just didn't want to waste a bunch of money for a failed print and to make sure I had everything I need to do it.
https://shop.prusa3d.com/en/prusament/1258-prusament-pc-blend-natural-1kg.html
Here are some good ideas for enclosures...they will help to prevent the crazy warping of your prints!
Thanks I just got rid of a couple of old tables just like these. I don't think I will do the tent idea because you can't see the print.
The tent is a bit weird...the table idea is cool though!!
What about the 3D printer enclosures that are already available. They are probably cheaper than what you would spend on the IKEA box in the video above. They are portable as well and take up less workshop space.
https://www.amazon.com/Creality-Enclosure-Temperature-Soundproof-Dustproof/dp/B0852TT63F/ref=sr_1_5?crid=192PLBHEPZPI0&dchild=1&keywords=3d+printer+enclosure&qid=1615326083&sprefix=3d+printer+%2Caps%2C241&sr=8-5
I was kind of thinking I would go with a clear plastic garbage bag and some dowel but look what I just found in the basement from when the kids were little, and it glows in the dark
FOC bake off. (From the Discord channel) . O'Drive, Moteus and MIT Cheetah
That was a good one...but all of those have a pretty high pricetag! I am also looking into Tinymovr, could be a good one? But both Moteus and ODrive are top choice boards!!
Found this on the web while browsing. The poor mans robot dog!
Link Here
Found this on youtube. Strain drive. It's a variation on the harmonic drive. The video at 14min is pretty cool to watch. James Bruton also has a similar video this week.
This came up in my Twitter feed. Apparently you can get an engineering addition of the alphadog for $2400 although price is over double on their website. http://www.weilan.com/en/alphadogc.html
You might want to scrub through this, but we've talked alot about this dog here. Here is a demo and unboxing. Very impressive speed!
Ok, I want one. Time to take out a second mortgage to pay for it. I wonder if it has sensors on it to keep it from running into my in ground pool. LOL.
I have found there is a big difference between wanting one and actually being able to buy it for advertised price.
Yeah i think they will eventually settle on 10k -ish...especially with the hype.
$10K sounds about right. Seems reasonable, If its something that will be useful all the time. I wonder if it makes a good guard dog?
Heheh yeah no dog food or p**p to clean up!
I guess it gets down to functionality versus cost. The unitree robot only works on relatively flat surfaces. if there are stairs then it will fall down. It can barely handle walking off a gutter based on the video's. It only has one front facing D435 sensor and short legs so I can't imagine this would be resolved soon. If it can't handle complex terrain, I will be honest I don't see a lot of advantage to a legged robot that can't handle stairs. My 3D Printed Dog can't get over anything higher than a couple of inches, and a well designed tank could easily handle that.
For some smaller projects. 5 amp limit..and based on ODrive. Fairly cheap at $43 per channel.
https://shop.m5stack.com/products/odrive-13-2-module-drv8301dca
I thought I'll bring the thread back to life with some progress I made on my actuator...