Sonic the Hedgehog Balancing Robot #2 : ELECTRONICS

[@jamesbruton]

This series is more R&D for load cells and CAN Bus - so I can apply those things to openDog if it's successful. This one will probably be a 3-part series.

[@DanielBillingslea313]

Awesome! Open dog is my favorite project

[@BoneKrusherKate]

Hey James Bruton I know you get this question a lot and if you do sorry that you are getting it again!. My question is what 3D printer are you using to print all you're modeling from? My follow up question is I am getting into Robotics such as Robot Dog, Robot Spider, Robotic Gripper (Hand) and even Robotic hand, my question to you is what 3D printer would you recommend for someone that is starting out to help with the Robotic building within the 3D printing world! (If anyone else can answer I do greatly appreciate it!!))

[@ZG-MC]

Hey James this is awesome what you are doing and fascinating! Just a suggestion as some of us do not have recources or money to build a robot. So how about a video showing us how to build a robot using cheap parts and no 3D printing. Maybe a budget £60 or under? Sorry for being super specific.

[@MikeCale]

Hi Jamies this project is amazing which CAN bus transceivers are you using? are they I2C? thanks

[@nathanchalecki4842]

Hi James! See the comment from the guy who needs help building a wheel chair?! Would be a perfect project for someone of your skills. Self balancing two wheel chair??!! I know reality and logistics (etc) destroy good ideas pretty quickly, but imagine being able to do something that really actually helps someone.. signed, your idol, nathan. Lol.

[@DominicGiles]

Great job James... What impresses me is the lessons you've learned in your journeys... Each robot is more capable.

[@madome3117]

Wow. This robot looks amazing. You really nailed it. Especially with the pid tuning and by how organic it responded to being pushed

[@jamesbruton]

thanks!

[@bruceneely4859]

this looks like a great R&D project, you should use it for testing sensors and navigation as you mentioned, it looks like a faster development platform then opendog

[@mvadu]

The ease in which he makes a standing robot balancing is simply amazing. 10 years ago how a Segway was out of SciFi and today a maker doing similar shows how far the hobby grade technology has come..

[@klausnielsen1537]

You are so awesome. When you pushed over that rampe repeatedly I was amazed at how alive it seems. Great demo and fantastisk progress.

[@nicholasjacob3594]

you can potentaly replace the 5mm pitch ball screws with 10mm in open dog to make the robot more dynamic

[@marco_gallone]

Not by far but certainly my favourite project yet

[@Mikesukes]

James you gotta be the most ingenious person on KZbin, always coming up with something remarkable!

[@jamesbruton]

thanks!

[@SnorwayFlake]

James to get the perfect balancing angle for a self balancing robot, hang is by the drive shafts from the axial so they are at the top of the robot, read of the angle from the angle estimator, add or subtract 180 degree depending on the range(you get the idea), then you have the angle for it to be upright. Yeah i know it'll change for this bot depending on the angle of the legs, but it is fairly simple to do, and saves a lot of guessing work.

[@MuditGupta89]

Watching your projects is always such great inspiration! Great job as always!

[@jamesbruton]

thanks!

[@theburntcrumpet8371]

This is the first time I’ve seen one of your videos. Really cool to see real robotics on KZbin

[@bryanjcera]

Shouldn't the leg going over the ramp be the one that "shrinks"?

[@jamesbruton]

Ideally, but that makes is oscillate when it comes off the ramp - although I have a solution for that in future episodes. Ultimately I want to test the loadcells out which work like suspension.

[@DerIstDerBeste]

@@jamesbruton I think if you drive in curves the amount of leaning should depend on the driving speed. I think the formula for centrifugal acceleration is a=(ω^2)r=(v^2)/r. So the angle of the robot should be α = arctan(g/a). If you only use the Feedback of the load cells you'd ideally get the same results as with this formula and the inertial measurement unit. But as you've found out it's hard to get an equal output from the load cells. So maybe you could use a combination of both data for a more reliable result. For example maybe you would get better results if you would only use the first derivative of the load cell signal and otherwise relie on the imu (PID to get level+curve angle which controls another PID which tries to keep the first derivative of the l.c.s at (zero + control input)). Or you could use the imu to 'calibrate' the loadcell scaling in your software. Btw when you use first order filters I often get the idea that that's a shortcut to solving the real problem. They make your robot react too slow, especially for something like driving over the board, which needs instantaneous reaction. Maybe you'd be better of with a third order filter for this application (the higher order will essentially lower the delay between the input and output of the filter). If you combine that with a low filter count your results should be much better. I think problems with oscillations should be fixed with higher dampening values.

[@iyaunaspan3915]

Bryan Cera You’re weird

[@nathanchalecki4842]

The load thing wont work during corners either. The outside leg will load up and shrink making the robot lean outward rather than in.

[@santinogaretto160]

This guy is a total genius.

[@mSparks43]

you should set up a store to sell the electronics and other components you use. No better advertisement for a product than seeing it working.

[@OzAndyify]

I'd buy a kit from James to support his work. This one looks like a keeper to me.

[@Getwright-]

Best Mecha Sonic design ever. Who knew the robotnik of our dimension would seem like such a cool dude. (Even if he did build Ultron and try to shoot a fellow youtubers eye out ;) )

[@nijram15]

It's really looks super professional!

[@DoRC]

I think you'll need the imu as well to basically make it lean into (not away from) turns like a motorcycle. Just relying on the load cells as the outside leg experiences more weight in a turn it will collapse causing the robot to tip the wrong way and fall over.

[@jamesbruton]

Yes, but I have issues with oscillations when it comes off the ramp - wait for part 4!

[@nilok7]

This reminds me much more of Silver Sonic/Mecha Sonic. Both of those designs uses wheels located in their heels to propel themselves at high speed much like your balancing robot.

[@apbosh1]

Wow that moves really nice!

[@Will_Huff]

New favorite robot! Great Job. Looking forward to this robot in the wild!

[@Innomen]

That is Extremely cool. This is my favorite thing from the channel so far. I'd love a wearable version of this.

[@janhetjoch]

Don't you want the leg thats going over the rampto shrink so the robot stays upright?

[@madome3117]

Totally agree. Maybe as an addition to the current version. It could give the whole robot more stability.

[@jamesbruton]

It's coming in a surprise part 4

[@timothyt.82]

@@jamesbruton It's not much of a surprise if you announce it, but since no one really reads the response comments, it shouldn't matter.

[@Humanoide33]

Very good start but you would get better measurements if the load cells were fully in the load path; they get very little strain where they are currently placed. Alternatively, you could stick strain gauges (with appropriate adhesive) directly on the aluminum extrusion.

[@martinedelius]

I understand that it's very convenient to have your workshop in your house but you are running out of space quickly. Imagine if you'd have 300 square meters of workshop to build in. :) Also, for shots where you want to show of details on the robot - like the motors spinning - try attaching an action camera directly on the robot.

[@jamesbruton]

yep, it's all coming up!

[@newburypi]

Thanks for yet another interesting video.

[@ahbushnell1]

I'm impressed.

[@themagnety]

Hopefully this comment doesn't get lost to noise. Not sure if you'd be familiar because off your location but Rev Robotics and Vex Robotics both offer brushless motor solutions with integrated encoders/hall efect sensors, and motor controller solutions. Issue is that they're built for highschool robotics competitions so the software may be difficult to wrangle because of that, but all of these solutions support CAN natively. Just thought I'd drop in, big fan of your robots, hope I have the budget to play with these toys like you do one day.

[@OzAndyify]

Looking really nice. I wish I'd built one of these instead of trying to build a walker (great for developing humility, not so good for interesting function). I reckon you will have to link the suspension to the IMU eventually, to handle traversal of a slope where one wheel will always be lower than the other, otherwise, you are limiting it to flat, level ground and very small bumps.

[@UnexpectedMaker]

James, you are too damn clever! Honestly, in awe of your knowledge and skills. Great projects, and fantastic explanations of how they work.

[@richardboyce4921]

This has turned into a very good R&D sandbox. compared to openDog you could say this is a RAD platform.. keep up the great work mate

[@jamesbruton]

Thanks, yes I'll be using this robot for ongoing R&D

[@ukulelefatman]

Fantastic. I admit I understand about .001% of what your doing, but still enjoy the hell out of it.

[@jespersolost]

If there was a real life Tony Stark award I'd nominate you. Great project!

[@dynorat12]

Looking good

[@tellingyoustories]

This thing looks great!

[@Karlemilstorm]

Hello James Very cool. I think accelerometer stabilisation was a good idea although the set point should be keeping the robot horisontal. Combining that with "softer" legs by amplifying the signals from the load cells could give some pretty good results I believe. Looking forward to seeing more

[@Karlemilstorm]

I believe the soft legs with your first order filter will remove the oscillations you were talking about

[@jamesbruton]

It's coming up in a surprise part 4 - the issue with keeping the robot level is oscillations when it drops off the ramp

[@foldionepapyrus3441]

I agree that is a working solution - Will certainly still need some accelerometer data even if the dynamic motion is defined by the the load-cells alone - what happens when the load cells are pushing the robot beyond tipping point in reaction to the ramp? Which with the speed this thing looks to have could happen quite easily I'd think - So will want a max tilt of the robot as a whole limiting the movement once it reaches that point. Though the filter may prove enough (for ramps it can traverse without tipping) - and adding in further filtering so traveling at high speed reduces the load-cells effect could also help).

[@simon-pierredandrea2928]

Cool robot

[@whatthefunction9140]

Well done sir

[@sergioa.1530]

Amazing job on stability 👌🏼

[@TiagoTiagoT]

But by removing the IMU for the side-to-side leaning, it will lean out in a curve, which is the opposite of what you would want...

[@tiporari]

Cool project and progress. Could you keep a running log of the encoder counts and estimate how far the bot has traveled? Might be fun to estimate ground position using relatively computationally simple subroutines. Could permit or help return to home or position holding accuracy (with a shorter buffer).

[@jamesbruton]

Yes I'd like to do that in the future and ultimately make it driven by position instead of angle/velocity

[@ovidiurosu6632]

Amazing, thank you very much

[@akshaykumargautam5289]

Showed up in my recommendation, amazing work✌️

[@Convolutedtubules]

Boston dynamics and agility robotics are cool'n'all but i'm scared of what you could build.

[@KilosFoxo]

he can build metal sonic

[@KilosFoxo]

@Shank Adams you could have said up N down N all around

[@andreyrumming6842]

Awesome project, but don't think we didn't hear the motors rev up at 5:51 when you put it on that chair! :)

[@jamesbruton]

I have a motor enable switch!

[@stuartpratt3662]

james if you have not allready gotten tired of hereing this you could simulate a hill (from side to side to test the load cells like you want to)

[@gooddaytodiy1349]

Thank you really helpful

[@daxx2k]

Awesome project! I feel the wheels white rubber are compressing a bit too much, don't you think? is that making the balance less predictable or it is the opposite because they ”grab” more terrain in this way?

[@lingxuanzhao7172]

Wonderful video！However I still have a question :Is it critical that the robot's center of mass and wheels are on the same vertical line？This problem has troubled me for a long time.THANKS very much.

[@DocMicha]

Extraordinary. How many hours did you invest in the whole work?

[@jamesbruton]

2 weeks so far

[@Stargazeo]

How can I learn about robot and make them and code

[@StevenIngram]

Wow, this project has come along in a hurry. (At least from an outside perspective. :))

[@StevenIngram]

PS. Very responsive bot!

[@TheAstronomyDude]

I found a load cell at the dollar store the other day inside a handheld luggage scale. The device itself was badly programmed and displayed the wrong weight so I got my $2 back but I kept the load cell :D

[@toakrikitt]

missed the opportunity to call it -Mecha sonic- but awsome as always

[@miwoj]

But can it go FAST?

[@jamesbruton]

wait for part 3/4

[@JCisHere778]

I seems like most parts of your controll-structure are somewhat "guessed". Why don't you parametrize a proper digital LQR based, digital, mutlivariable control? I think it would greatly improve both stability, dampening and get rid of negative feedback on the individual control-systems. It would also get you much more positionally accurate. (The oscillation when it was supposed to be standing still. I know there will be some left because of backlash etc.. But still, that's quite a lot)

[@naumanshakir9344]

How do you manage to make robots on this much big scale? Do you make things yourself or do you have a team? I'm curious to know.

[@KacperLaska]

Great Job but I have a question about the delay between the leg sensing deformation to the time it takes for the controller to compensate, is it the first Order filter that’s causing that ? It seems like that delay might be problematic.

[@jamesbruton]

Yes that is smoothing out the reaction, but otherwise each leg affects the other - you can already see a lot of jitter in the actuators.

[@soccer2cubeing]

Which 3D modeling software do you use James? I don't recognize the interface.

[@xwinglover]

I love sonic

[@remcoblom4047]

Wich encoders do you use and where can i find them

[@OfficialRetroPlayCast]

Metal sonic!!!

[@1nickyross]

Are you still doing the gonk droid? (Finishing it off or something like that)

[@bukachell]

YES

[@gistnoesis9116]

Well done ! But can he jumps though hoops, and do the looping :)

[@jamesbruton]

wait for part 3!

[@OzAndyify]

@@jamesbruton Ooh, jumping! Bring it on!

[@neihmar]

amazing project !! which 3D Design software are you using ?

[@neihmar]

Fusion 360 :) just found it thank you

[@jeremyduhamel6754]

Un génie !!!

[@circuitdotlt]

Damn. Is this your personal project? I mean, things like this need a ton of time and it is hard to imagine you manage to do all of this after your day job. Or is this part of your income?? Anyway incredible.

[@circuitdotlt]

P.S. I think you still need some kind of non-oscillating mechanical spring to quickly react to sharp bumps. If you know spring ratio and load, you get compression.

[@Jigzbo]

Just wondering, is there a part list for each project?

[@ChrisLocke1969]

I don't quite understand why you're shortening the opposing leg to the obstruction. I would think the one with the obstruction should shorten, as we would bend a knee to take a step onto a stair, "shortening" the leg with the obstruction. No?

[@jamesbruton]

It's coming up in a surprise part 4 - the issue with keeping the robot level is oscillations when it drops off the ramp

[@ChrisLocke1969]

@@jamesbruton great, thanks, i can't wait for the next video... I love your projects so much! 👍

[@busiefurball3917]

Will it go fast?

[@jamesbruton]

eventually

[@clonkex]

@@jamesbruton Looks like it could already go pretty fast if you had the room do try that. But no idea if your code is smart enough to prevent it going near the maximum speed of the motors (which of course would remove any ability to balance, which is OneWheels have the pushback feature).

[@mohitsilori6064]

wow super 👌👌👌🔥🔥🔥🔥🔥🔥

[@harryvivian8403]

What software do you use ?

[@BoneKrusherKate]

Hey James Bruton I know you get this question a lot and if you do sorry that you are getting it again!. My question is what 3D printer are you using to print all you're modeling from? My follow up question is I am getting into Robotics such as Robot Dog, Robot Spider, Robotic Gripper (Hand) and even Robotic hand, my question to you is what 3D printer would you recommend for someone that is starting out to help with the Robotic building within the 3D printing world! (If anyone else can answer I do greatly appreciate it!!))

[@homerobomaker9074]

Why r u still using MPU6050 and why not MPU9250

[@Gualor_]

Hi guys, i just wanted to ask to some engineers questions regarding the control and stability of robots based on MPU measurements. I know that in control theory PID controller are very useful in the making of a feedback controller with black box model (just need tuning of the parameters), but in the case of simple mechanical systems such as this one, wouldn't be much more efficient if we could just measure mass, moment of inertia and barycenter of all rigid bodies in the system and calculate the torque needed to maintain the dynamic equilibrium?

[@foldionepapyrus3441]

You are effectively solving the same problem in much the same way - the tuning of PID values to get the response you want is the in the real world how does this mechanism perform iterative version. Where you are describing a pre-calculation based on the mechanism properties it sounds like - which could be a very computationally efficient way to run it but in practice will be much harder to make work - the PID loop tuning process accounts for production tolerances and as it needs to be a high speed loop to work at all will likely work better at adapting to unpredictable noisy input. (Unless I'm not following your meaning at all)

[@Gualor_]

@@foldionepapyrus3441 yes, you have been very helpful and you did get exactly what my doubt was. Im just guessing here, but maybe more complex behaviours, for which the delay of the convergence of a PID controller would be determinant, having a good feedforward controller based on the real physical system could improve the overall performance. Does It make sense?

[@foldionepapyrus3441]

@@Gualor_ I think that could work well when there is not enough sensing for precise mapping of the whole mechanism or computing cycles are too slow to react. For example Open dog itself having more legs could perhaps make use of precalulated responses to work effectively with no IMU at all - just load-cell sensing on each leg to know which foot is down under what load -and effectively infer all the torso's IMU info from this and known geometry of the limbs. (though in practice to walk it wouldn't do the inference - just lookup the next correct response based on input of wanted movement direction and load). Or in the case of the PID loop taking too long the lookup based response could be implemented in the leg itself, but able to be overridden from the central brain if it does the wrong thing - a bit like our nervous system flinch reflex.

[@notbeta1264]

You should try to build a Chuck E. Cheese 🧀 stile animatronic

[@oscarr2k8gaming68]

Hi James I remember coming to oli b school about 4 years ago with the bb8’s

[@Bruce.B]

So, it is Mecha Sonic

[@fryingpan37]

44th!

[@Scott_C]

Mount a Google Home tethered to a wireless hot-spot and some active object tracking and you've got a semi intelligent droid to follow you. 😄

[@TheBrickLot]

Least he's adding these fancy legs on Sonic and not his Ultron

[@shaqirshahini3130]

I prefer the head looks a lot like Sonic's actual head

[@apershin77]

could you please try to make it fast

[@TheTruth-ui1cm]

How about making a fx-7 medical droid from Star Wars

[@jim2386]

Dampened - to make wet Damped - to reduce oscillation Please don’t dampen your robot....it won’t end well ;)

[@fryingpan37]

isn’t it the other way around...?

[@jim2386]

uno reverse card nope...it’s damping ratio, not dampening ratio.

[@TATTIEPICKER]

You must be fun at parties.😬

[@jim2386]

phixion. You know it! :)

[@Chokun-Gaming]

So... Real life metal sonic?

[@animateorelse1696]

Zooommmmmmm

[@MrRohJebat]

@james bruton

[@TS-kg4lf]

Kind of worried to see u using 2 batteries without BMS or any monitoring

[@jamesbruton]

they are charged with a balance charger

[@wendycato785]

You can put fome feet

[@natansoares9409]

what engine do you use?

[@Nialon-The-Spy]

Make it look like metal sonic it makes more sense that way

[@Kyusoath]

If you add arms please remember sonics arms aren't blue.

[@sanlin_Xavier]

Can you mentor me bruton

[@quickend666]

I am not saying you are evil just that you are definitely building mecha sonic hopefully not for evil

[@hungnguyen-yn1io]

Good.my name is minh hung.hello

[@starved6493]

What if it fell all apart and he got mad 🤣

[@DevEncryptionNull]

openDog openDog OPENDOG!!! OMG WTF happened to openDog?

[@jamesbruton]

This is R&D for openDog as stated in the video

[@honaynonay2631]

bender's face 😆😆