A droideka build from Star Wars would be so cool and I feel like it could really fit the style of this project too

@glowytheglowbug

2 жыл бұрын

Yup

@lr5612

2 жыл бұрын

That would be awesome!

@Tomd8002

2 жыл бұрын

Was just about to say the same! Great minds think alike!

@MrJodiH

2 жыл бұрын

I had the same thought. Also reminded me of this guys ball robot kzread.info/dash/bejne/q6JnqLmbXdm3gZc.html

@Unmannedair

2 жыл бұрын

That shouldn't be t to hard to build. You can steer and propel it using rotational moment and then have legs for turning by way of standing or cg shifting.

Looks like it would require a 'brain' to decide when to morph into the next shape so as to catalyze movement. currently it's like being on a swing and pumping your legs in the middle of the arc instead of at the ends.

@mikeloeven

2 жыл бұрын

Installs "Brain" :What is my purpose? :You're a ball. :Oh My God !

@planetdesign4681

2 жыл бұрын

Yeah I feel like it could utilise some sensors and an ai as to know when to fire each servo, because you can see it just about to roll over but then another one of its servo’s push it *backwards* instead of forwards.

@tbuk8350

2 жыл бұрын

@@planetdesign4681 yeah, that would likely be easily handled by a simple inverse kinematics algorithm and an inertial measurement unit like they've used on their other robots.

@alejocouderc4495

Жыл бұрын

I would like to be brain of the ball robot

James: There is no stationary reference. We need an inertial measurement unit to help measure the angle. Me: Ha ha look at the squidgy robot ball!

have a look at Zenta's MorpHex robot, does a simmilar method of movement but with each eighth (quadrant???) of the sphere being separated.

@griffinoconnor1523

2 жыл бұрын

Had the same thought!

@Fredjikrang

2 жыл бұрын

Love Zenta’s stuff!

Re: Ghost cubes. Is it really a big deal to have the cubes touch each other? Could there be some progress if the cubes are exploded a bit relative to one another to allow room for the pivot?

@doxielain2231

2 жыл бұрын

Or if the cubes are frameworks with internals?

@gearandalthefirst7027

2 жыл бұрын

chamfering the hinge edges could also give room for the hinge structure perhaps?

9:25 this is the point where I realised it would have been so much easier to just write a machine learning regression algorithm to, enable the ball to learn to efficiently roll on its own, than to program it manually. Seriously consider this, it will not only save you a lot of time and efforts, but the ball would actually learn much better than you can ever hard-code from your limited human knowledge, logic and imagination. Because the complaint mechanism you built certainly had the capacity to roll itself perfectly in any direction, that too smoothly, with perfect enough actuation timing. It's like teaching a robot how to balance on a cycle, probably hard to hard-code but a cakewalk to let it learn on its own.

@Shaurya_Pant

2 жыл бұрын

Maybe add a gyroscope (which again is optional, as the ML algorithm can just figure out which way is up from the visual markers that you can put on different sides, and let it see itself roll. Or you can talk a more hardware orientated approach with accelerometer and gyroscopes, and let it learn internally without using a camera and sensor, but both will produce a perfectly rolling ball that you can even control with a remote. (Setting the end point, and your system eill actuate itself to that point)

@Charles-Darwin

2 жыл бұрын

Do you know of any videos or articles on this approach? I would like to learn about this myself

@thijsvanliempd2824

2 жыл бұрын

it's not difficult to hardcode at all, each (euler) angle from the gyro/accelerometer directly corrosponds to servo values (just the sum of an offset/phase-shifted sin() multiplied by the range of the servo), to give you a different resting orientation. To make it move, i'd guess to just add some more phase shift, so the natural resting position is just off (in the direction you want the robot to move). Right?

@xandermasotto7541

2 жыл бұрын

Even if it's machine learnt, he's gonna have to add IMU measurements anyways in order to compute a reward for the RL agent.

@europhil2000

2 жыл бұрын

I certainly don't want to talk trash on James because this is a great project. But have a look at the code on github. It is literally just moving the servos to some predefined position and repeating regularly. The code is less than 75 lines long (incl. blank lines)! So I guess before throwing ML at it he'll probably add a bit of dynamical model first...

This reminds me of one of those frisbee balls from the early 2000s

@clonkex

2 жыл бұрын

I had forgotten about those! What a blast from the past

@Matty.Hill_87

2 жыл бұрын

My kids have one of those, they're out playing with it now 😂

@this_commenter_had_a_stroke

2 жыл бұрын

same

@rhiannonband

2 жыл бұрын

They've made a comeback. My son got one for his birthday.

I'm always amazed by your creations. Even one of your projects would probably take me a year to make!

@anthonyanth8368

2 жыл бұрын

@Gadget Addict. I know how you feel brother kkkk

My dude really just made one of those nerf collapsible footballs lmao For real though seeing your entire process for design and manufacturing is really cool

I love james recent content. The prototypes are so creative and interesting and the choice in footage is honest and i love how he shows the entire process with successes and fallbacks together. Such an awesome lil robot

Maybe something with 3 phases (3 * 120°) would provide more consistent rotation ?

@torokati44

2 жыл бұрын

Or 4, cycling around in one direction, instead of just switching between two states, indiscriminately of which direction it's trying to actually roll.

@zaidro1

2 жыл бұрын

no necesarly is more important in this case program the ball whit some sort of retroinformation of the speed and posicion to know wat is the next step to move the ball

Have you considered using machine learning and train it to move effectively? Set up an enclosure and a sensor that tells the AI where the robot is in relation to a goal and let it teach itself how to move.

I love the noise it makes kinda sounds like ASMR when it rolls, Love your projects keep up the great work!!!

great to see another video by james

This has been done. There's a white ball that's rolls along pushing out the relevant middle joint at the right time to keep up the momentum. It can then transform into pentapod bot and walk. Let me know if you want a link to it. Great work, otherwise. You're on your way to converging on the existing one.

@cameronbest

2 жыл бұрын

Can you post a link to that rolling robot?

@mozkitolife5437

2 жыл бұрын

@@cameronbest MorpHex. It's been a while... It's actually a hexapod. kzread.info/dash/bejne/aq2Dj6prYsi0fso.html

@HaloWolf102

2 жыл бұрын

Sorry, you are mistaken. The robot that James has made, and what Zenta has made is not exactly the same. Zenta uses the method of 'pushing' off the ground with many different 'arms'. James uses a mechanic where the robot is morphed to an extent where the center of gravity is always infront of where the ball wants to go, but it does look very similar, and I understand why you thought they were the same.

@mozkitolife5437

2 жыл бұрын

@@HaloWolf102 You're right about the differences in locomotion but I still consider them both as locomotive spherical robots. Obviously not 100% identical. Just like two different tetrapodal animals that move differently. One way might be more energy efficient etc but they're classified together further up the binomial hierarchy.

@HaloWolf102

2 жыл бұрын

@@mozkitolife5437 Would you consider James's BB8 robot the same as this?

You've imagined something you know you can't build and prototyped it. Your content is an inspiration to armchair engineers everywhere... keep it up

@owenkegg5608

2 жыл бұрын

I tend to imagine things I know I can build, start building it and then hit a software bug, sleep it off then forget it ever happened and wonder why there is a vaguely phallic robot now on my desk.

The inspiration for this project reminds me of the old "Rubic's Magic" puzzle, a series of plates with images of rings, linked or un-linked, that you manipulated to make the rings appear to change from one state to the other. The edges are all linked with plastic filament and allowed it to flip and shift shapes.

3:04 "played around in cad"... didn't you already show your Cardboard Aided Design...? On a more serious note, adding an IMU to make it more dynamic seems like the next step. Right now the morphs sometimes push it along and sometimes seem to stop/slow it down. Also, the first cube hinge design you showed, with the servo doubling as a hinge would work. For the cubes to fold correctly they only really need to touch at the corners, you can clearance the edges/faces for the servos as much as you want/need.

@pacman10182

2 жыл бұрын

I see you watch bad obsession too

I like that you took us along for the ride to discover this robot design. Interesting to see and gives me a couple ideas too.

It's beautiful as always 💗

I cant wait to see more of this little ball!

@mrupsidedown9879

2 жыл бұрын

It's kinda big

The thing that crossed my mind when you were talking about the connection between blocks was muscle wire.

Excellent as always James :)

Congratulations on 1milliob subscribers keep making us inspired and think creative thank you love ur vids ❤

Interesting to see the interplay of processes here between you as an engineer (albeit with aesthetics as a consideration) and Erik as an artist (albeit with engineering as an interest). An actual collab would be fascinating to watch.

Zenta's Morph Hex is still amazing kzread.info/dash/bejne/q6JnqLmbXdm3gZc.html Maybe there is some inspiration here.

Excellent project 👍

Pretty cool concept, I enjoyed watching it :)

James! You just need to add some inertia to the middle of the petals to get it to start on its own, take some steel bars and strap them on the inside petals.

I think it's funny you build robots every day but spin the display with your hands lol

@PinataOblongata

2 жыл бұрын

Simplest solution to the problem ;)

Congratulations on your 1 million subscribers 👏👏👏

Adding to the electromagnet comments, could you have free pivots and activate/deactivate them to lock faces in place? So instead of actuating the hinges you can make areas flexible or rigid

did you ever consider engaging all the servos? Having 2 adjacent side squish while the 2 opposite sides expand? seems like it would roll on 2 edges rather than the face. when it rolls on the face, it is also balancing on the face. You achieve a significant amount of balance because of the symmetry, but variations in the individual parts could throw off the center of balance.

@marcus2of10

2 жыл бұрын

I think Les has the right idea. Two on one side stretching and the opposite two compressing. No design change required. To make it roll you need to shift the center af gravity away from the sphere center. Then at the appropriate time return those servos to neutral and get the next group working. Maybe a gyro would help.

Super cool!

If you're going with a sphere; there's a design I saw a while ago that was a twin-hexapod (basically two hexapods merged at each-other's back), which had legs that had the outer surface of the last segmentm, the "shins", shaped like a half of an orange's, uh, forgot the word, the wedge shapes you can separate inside the orange); so that when it would pull it's legs in the tips would join at the poles, and the tops of the hexapods' shins would meet at the equator. It was able to move both as a legged robot, but also by slightly pushing the legs out of the spherical formation it could roll along the equator, and turn by doing it asymetricaly between the two hexapods.

I liked your ghost cube concept of putting servos in each cube, and using strings to pull them together. It occurs to me that to over come the cabling issue you could have the power/control cables run inside the strings you use to pull the cubes together.

I love oyur videos James. You are willing show show things even when they are less successful or even complete disasters.. :D Thanks. BTW I think the key to getting like this ball to work is resonance coupled to precise feedback.. The same basic problem with all walking robots.

This would be a great candidate for algorithmic learning, given the speed/timing/directional response interplay, so many variables!

You are welcome for the 1 million subscribers:)

As an open loop automata it does pretty well. Closing the loop with an IMU would be fascinating, ,love the linear style drive and topology, The right SW could really make it move I expect. Great work.Looking forward to see how you translate , like some sea urchins ? penta

Holy crap just realized you made it to 1 million subs! Are you going to make a celebration Video? Congrats!!!! Anyways, love these concept models and working through their evolutions!

seems like it needs an angle sensor, it's being driven synchronously when it needs to morph based on its angle

Watching you goad these small bots on the floor reminds me of a proud parent helping a child to walk.

It's basically Magikarp in ball shape... Waiting for its evolution into Gyarados

Isn't there a similar shaped toy, only instead of a robot it a frisbee that turns into a ball?

@HollowRoll

2 жыл бұрын

I was thinking the same, though I can't remember at all what it was called Edit: it may have been the "phlat ball." That's at least the same sort of toy

@pacman10182

2 жыл бұрын

@@HollowRoll that's it, spelled with a ph because we haven't progressed from the 90's

Really impressive! Why not considering magnets for the cubes ?

I'm waiting for the moment when he becomes to Dr. Robotnik/Eggman. Another cool project James.

@Matty.Hill_87

2 жыл бұрын

He's one good day away from becoming iron man.. But also one really bad day away from becoming a super villan 😂😂

This looks interesting, you should definitely try out reinforcement learning to create a controller for it!

Can you use electromagnets embedded internally on the sides of the cubes to push and pull the faces of the cubes?

@chrisk8208

2 жыл бұрын

My thoughts exactly. They'd also provide the pivot of a hinge.

@GrantStephens

2 жыл бұрын

Thought the same thing, but rare earth magnets on a servo that moves it to and from the face of the cube

@nou4898

2 жыл бұрын

everything good until the thing gets stuck to a metal surface

I always enjoy your videos James. I’m just never sure why 😜

Cool concept! With the Dune movie coming out, I think building some kind of morphing worm robot would be cool... Or an ornithopter, but I guess it would be quite hard!

We’ll done I just realised you had 1m subscribers it was a fun journey to get you where you are but keep up the good work

Intresting concept with a more sophisticated controlling software and sensors as you said

Zenta's MorpHex has always been a big favourite of mine

Awesome

This is a very interesting idea, it reminded me of that bowling ball with Mark Rober. maybe try adding an inertial measurement unit so it can better time when to morph. idk.

I remember seeing a hexapod that can fold up into a ball a few years ago. Would be neat to try and replicate that idea

@mozkitolife5437

2 жыл бұрын

MorpHex

What you could do is add a mechanism that can detect rolling and sync the motion of the servos with it's rolling speed so that the servos in the direction in which it is rolling will always contract and provide force in that direction at all speeds

I wonder if you could use a gyroscope at the center of the ball to measure the angular rotation rate when it is pushed? If so maybe through some machine learning you can have the arduino control the arms through a changing tune PID protocol to flex and retract the arms at a frequency that corresponds to how fast the ball is rolling.

Wonder if electro magnets would work to push and pull the sections part and together

This guy is a genius!

Driving the cubes from a central motor could work if you have a thin link that connects the two cubes. The link might need to be-able to stretch as the cubes rotate over each other, but it could work.

Could you use electromagnets to work the hinges? The hinge is attached a flexible joint and inside the box, electromagnets either repel or attract in order to drive the joint into the correct orientation (clarification: I'm talking about the joined boxes at the beginning of the video)

Just a quick thought of me, instead of using servo's cant you use electromagnets inside the cube? By attracting and propelling you can probably move them. And contact points on the surface of the flat can provide power, only problem that it is lost during the movement so there has to be a small battery inside to overcome the loss during transition. Great project with nice challenges.

Awesome video, thanks. For those wanting to skip the sponsor 7:08 to 8:16

An alternative to an imu might be capacitive sensors, because it's so symmetrical and omnidirectional, all you really care about, is what sides are touching/close to a surface

How about using electromagnets you can switch between repulsing and attracting in different faces, to drive the folding from inside the cubes? (maybe locating them closer to the edges, so that different edges can have different polarities when necessary)

Very interesting (as always). Maybe a bit of weight in the middle of each of the four segments - that would give a bit more of an asymmetric balance force?

You would need to add a giroscope and program it to shift in order to always keep a lump behind it and move that way instead of in a timer

About a mechanism for the boxes- in most cases there are only 2 transformations- means 2 states. That could maybe be achieved with electromagnets inside the boxes- so corresponding surfaces can repel and attract on command but still can only go in the right way because of a thin hinge like in the cardboard models. It would be a very fast an brutal transformation though

Crazy Future!

This really reminds me of the 4 legged robot ball from Carl Bugeja. Maybe there's some room for inspiration there, a lot of the problems you mentioned I also remember hearing in their video

You could maybe use electromagnets to move the cubes with a really thin link

Can a distribution/control of the mass center on a ball-shaped robot be an easier solution. Sort of what is is used in toys for cats.

tendons attached to motors in the cubes?

Would it possibly work better with a more sinusoidal transition? The movement of the servos seems a little binary, so there’s almost like a bump that it has to roll over before changing shape. Really cool design, love watching the process!

Mass man Vere level nenachadu lam panre 👍👍👍👍👍👍👌👌👌👌👌👌👌👌👌👌👌💐

You could make like a sphere of actuator arms and code it to extend one arm in the direction you want to roll then pull it back in before it hits the floor and then have arm above follow and so on.

(So I have the urge to comment this to every of James' video, but) WHERE DOES THIS GUYS TAKE ALL OF THIS TIME TO MAKE THESE EXTENSIVE PROJECTS!!!???

Nice

An interesting way for movement can be magnets attracting and repelling each other

Did you do split testing to determine talking in that weird cantor improves views?

james you have to design in a way you can change the center of gravity then it wil move toward it...

@kasuraga

2 жыл бұрын

this. you have to factor in the center of gravity more to make it roll. It'll definitely have to be able to control the rate of motion, vs being a static cycle.

@banaankiamanesh

2 жыл бұрын

@@kasuraga right. And corresponding to the distance of COM and center of the ball youll get speed. if it can be modeled ...

Would be cool to try a ball robot with pegs that push out in whatever direction is needed to control motion

You could design the cubes to have interlocking hinges built into the edges of the cubes. a metal axle could then be slid in (like the pin in a hinge). one cube has the drive motor and drives the hinge pin/axle. this cube must be free floating on the axle so the axle can spin. the second cube would be fixed to the hinge pin/axle, and would then be driven by the internal motor inside the first cube.

Impressive .. you should try doing motion of snake of some kind of animal

The use of Nitinol wire joints would provide the thin edge connectivity and low current demand to adopt its memory form together with low weight

Instead of hinges it seems like some form of electrical muscle/smart material would be significantly preferable for the morphing shape - being able to act almost exactly like the thin tape. I don't even know if anything like this is actually available though, because the expanding/contracting mechanism in them would have to be extremely small and thin.

@chrisk8208

2 жыл бұрын

Perhaps pneumatic rubber tubing?

Reminds me of tumblebutts. Would love to see this turn into a droideka that a previous person suggested.

Definitely gonna need a couple gyros and a logic computer, but this is an awesome idea

I would be interested in seeing this concept applied to a rolling wheel or rolling pair of wheels. A cylindrical wheel would be easier to control than this ball shape, and you could reuse the flexing mechanisms from this design to make the wheel roll. Having two such morphing wheels might even let it turn if the mechanics work out the way I think they would.

I think if the timing was calibrated to it's angular position using something such as a 6 axis potentiometer you should be able to get it to flex in a way where it's timed correctly whether it's at the start or at speed, or for different surfaces.

@lio1234234

2 жыл бұрын

Oh, I just got to the end of the video where you suggested the same thing. Sorry 😅

If you were to try the cube structures again, you could use a servo inside the cube pulling on a ribbon like the Jacob's ladder toy. That would give you the (near) zero clearance hinge you were looking for

How about trying to make it do the same pattern but diagonally, morphing all four sections?

Slap an accelerometer in that bad boy to sense orientation and actively control which segments of the ball to actuate. Bet with some nifty programming you could get basic steering and decent speed. Regardless, great work and would love to see you come back to this project

Master, destroyers!

What about the corners are part of the hinge. Maybe something like this hinge on digi-key, but the inside mount has a gear on its internal that interacts with the drive from the other cube? (45 degree onto corner not off the 90. so the corner is chambered.) Digi-Key Part Number RPC2387-ND I am a huge fan of your work, keep up the creativity!

OMG This is AMAZING!!

@dhairyaagarwal716

2 жыл бұрын

does NASA use it in its satellites or something?

Maybe you can control this robot like a tank. Creating undulations in each plane that you can speed up or slow down through remote control. Then your ability to control the undulations in both planes may allow you to turn it as well while progressing it forward. Just a thought

The cubes might be realisable by adding electro magnets to each cube that push and pull on various sides. What do you think?

@tijnmestrom2153

2 жыл бұрын

I had the same thought. This is a great idea imo