A robot balancing on a ball R1.1 / 玉乗りロボット

[peiznkev000]

Have you guys thought of scaling down this technology and possibly using it as joint? I could see this being used to replicate a joint in a humanoid. Or scale it up to create joints for dynamic structures.

[makumagai]

@peiznkev000 I'm not, but many people thought that idea, and told me. There will be at least two challenge. One is how to hold and how to sense the rotation. Partially, I solved it; see youtu.be/qnH_zoAYDMg. The later is how to transmit the power to a smaller ball. The ball is driven by friction on the surface, which will limit the torque that we want.

Scaling up seems interesting. One of my friends questioned me whether I can drive a ball with 1m diameter holding a person inside:-)

[zooto68]

What is the ball made out of and what are the wheels made out of please?

[makumagai]

@zooto68 The ball is a 8 lb ball for bowling, coated with Plasti Dip spray.

And the main materials of the wheel is aluminum and duralumin. The roller is made of extra super duralumin (ESD), A-7075.

(The strength of the ESD is not needed but the shape of the roller cannot be turned when using aluminum due to its plastic strain, once we used brass)

[TheHomebot]

@makumagai, fantastic robot and awesome job. Are we able to get a hardware list? I am guessing the splendid control of the finale robot "A Robot That Balances on a Ball" is the code of course, but also the stepper motors you used to get the micro movements right? I am just a home robot hobbiest and love playing with new ideas. Again great job.

[makumagai]

@TheHomebot Thanks. I don't have complete list of the robot hardware because it is composite of several 'standard parts in lab' in many format data(AutoCAD, Eagle, OrCAD etc).

But a core parts description is in Fuji Technology Press /finder/xslt.php?mode=present&­inputfile=ROBOT002200030013.xm­l (require free registration). I have a plan to make Pro/E model of the robot but not yet. The code can be sent via e-mail.

[crossbowjapan]

The original Ballbot of CMU was merely a right-angled 2 axis balancer, whereas this one has 3 wheels enabling to pivot. What it needs is a good yaw gyro to keep track of where it's going, thereby implementing the real tranport solutions.

[makumagai]

@crossbowjapan Thanks. The original Ballbot has additional axis in yaw above inverse mouse ball drive. The problem on controlling our robot in yaw axis is to detect actual angle because gyro signal must be integrated, which sum up zero point error. In detecting vertical angle (inclination), it is compensated by accelerometer, but no good absolute sensor for yaw. Probably it requires localization using map and range finders.

[makumagai]

@crossbowjapan I know that one, because I'm now staying at Ralph' s lab:-) It's seems to be very expensive. This robot has sensors of 250USD in total.

And that reminds me where I saw your name 'crossbow':-)

[DonQuichotteLiberia]

Nice job! What type of controller are you using?

[makumagai]

@DonQuichotteLiberia Thanks! The controller of the robot is 16bit microcontroller Renesas H8/3052. It is one of most famous MCU of 25MHz in Japan. 

[DonQuichotteLiberia]

@makumagai Sweet, thanks for your response! Are you using a physical model of the robot, as in LQG or H_2/H_\infty control? Probably a PID controller is not good enough?

[makumagai]

@DonQuichotteLiberia No model, No LQG, No H\infty is required:-) Only PD feedback of sensor and estimated rotation of ball, and parameter tuning technique by hand (but only takes less than one hour).

Model besed control is sometimes powerful but sometimes only waste time, for this kind of simple system, I think.

[DonQuichotteLiberia]

@makumagai Awesome. Thanks for taking time to respond. :-)

[mremgee]

I've been trying to find wheels like that but all I can find are ones with two lines of off axis wheels. Irritating! I'd love to see some code examples on how you calculate engine speeds based on direction and such. Is that available somewhere?

[makumagai]

@mremgee I designed the wheel by myself under permission of patent holder RIKEN. You can buy original (larger) one on the net, though unfortunately they have only Japanese page. Probably, they can answer you in English by e-mail.

For code, I can send you the source as is. Please let me know your e-mail address via direct message to my account on YouTube.

[chattmataz]

very cool!

[makumagai]

Thanks!

[CYP2E1]

Hello,

I like the clean design of your robot and the stability you were able to achieve. My question is: what about the rotation of the robot? If it needs to turn around? Wouldn't a balancing bot on 2 wheels be mechanically sturdier, simpler, cheaper AND permit rotation?

[makumagai]

It's due to application of the robot. I think it's more interesting to turn and we didn't consider a designed that cannot turn when we planed our project.

In another question, yes of course two wheeled balancing is more stable, simple, easy to develop (so we also have severals), but it cannot move to side direction without turning. Imagine when you kick a rider of a Segway from the side:-)

Our robot, and of course pioneer 'ballbot', have advantage that it can escape to any direction at once.

[vandallas12345]

have you actually watched the video? look for "Pivoting" around the 2 min mark

[OutlawTomFantastic]

Why does the video act like this is type of robot is original and came from some Japanese forum? Ralph Hollis was the first to conceive of and create the Ballbot.

[makumagai]

Of course, ballbot is the first one that realized balance using a ball. Did you read description "like 'ballbot'" and date this movie appeared (which is later than ballbot) ? :-)

This one also balance on ball developed in Japan but is based on another type, without knowing ballbot.

In addition, I'm now visiting his laboratory to collaborate on these kind of works.

[juan1815214]

the web name is "Cosas de Mecatrónica"

[juan1815214]

I was to search videos in Youtube about robotic and I found this video so interesting, I decided to write a article in Spanish for a mecatronics web, the title is "Robot equilibrista" and I hope you like

[makumagai]

Unfortunately, I cannot understand Spanish (though I have learned French a little 15 years ago), so I will not be able to read your articles.

But, yes, you can write your article and can introduce this youtube movie , as some blogs do.

[alonohev]

That is butifull!!

What kind of controll did you apply?

Where can I find your paper of it?

[makumagai]

Simple PD feedback control on inclination and position.

You can find one of our paper on IEEE explore.

Please search 'balancing on a ball' there.

(ICCAS 2008)

[DirceuRodriguesJr]

Very good project! What was the processor that you used (Renesas based)?

[makumagai]

Thank you.

We used Renesas H8/3052 processor(16bit).

One of popular board by AkizukiDenshi.

[Mudskippersam]

Does it use a holonomic drive? If that is the word, it seems familiar to me!

[makumagai]

It's a kind of holonomic drive using three wheels for omnidirectional robots proposed by Riken.

[bobbin221]

Great Video! I like the usage of the omni wheels for the drive mechanism. What type of ball did you end up using, and where in the world did you find it?

[makumagai]

The ball is rubber coated bowling ball.

The ball is cheap (it is very important because of simple structure inside) ordinary 8 pound ball.

The rubber is Plasti Dip spray (google!) that was introduced in `niconico video as mentioned' in video. We have used that kind of ball till now.

[kipp8]

awesome !!! given a little more development this type of platform could potentialy capture the holy grail of robotic ability and climb stairs

very dissapointed though to see a teather if only for the sake of control. Why for the love of god is this thing not remotely controlled???

[makumagai]

> tether for control

The robot does not have autonomous decision making in it motion plan. It only can balance on a ball.

The controller is only for commanding start/stop of control, changing operation mode, or what direction to move.

(The controller will be wireless soon.)

In this sense, this robot is still a mobile platform such as wheel part of wheeled robot.

Autonomous motion such as obstacle avoidance should be built on this robot in the future work.

[lubbiz]

Awesome, but is's useless! Now guys, start making useful things for a change xD !!

[makumagai]

The next robots under development will be more useful, I think.

The video will be uploaded early next year, after we propose them in technical conference in Dec. Please wait it.

[lubbiz]

Really nice, really looking forward to it! Keep it up :)

[jpend1]

This sort of technology will have an extremely large number of uses in the future. What we have here is a device that can move an item (sitting at the top) with 3 degrees of freedom (translation in a plane parallel to the ground, and rotation about the central axis). It is compact and able to perform this motion with a single point of contact to the ground. One use I can think of is at a bar: the robot could be used to distribute drinks or receive empty glasses for transport back behind the bar.

[chattmataz]

@jpend1

I can already imagine drunk dudes pranking the robot :-D

[Gesthrel]

Very nice work

Good job ^^

:P

[FutureMachineLifeBot]

great job! i love the design, very clean