Rise of the Ping Pong Robots

mikejuk writes with this excerpt: "Meet Wu and Kong — the latest in ping pong playing robots. They may not achieve exciting matches at the moment, but the fact that they can do the job at all is an indication of how fast things are moving. Unlike many other game-playing robots these two players are humanoid and are kitted out in old style Chinese jackets. They are about 1.6 meters tall and weigh in at 55 kilos. They track the ball with video cameras situated in their heads and then play a variety of strokes. They were developed by Zhejiang University and are currently turning up on the Chinese media as a novelty item. ... The current record for a rally is 144 rounds between robots. Humans can compete against them, but the robots lack the variety of shots that makes table tennis a game of strategy as well as accuracy."

Comments

FYI

[Nyall]

The first video has some text saying that the current robot vs robot record is 176 strokes.

Re:FYI

[Anonymous Coward]

Let's hope we can get one of these robots in the shower with Jerry Sandusky and really see what they're made of

Re:FYI

[Anonymous Coward]

Even if you wanted to, there is no suitable safety towards monocle-inducing subplants.

Re:FYI

[Ihmhi]

ERROR: ball not moving

Applying additional force, continuing to swing.

Wireless Joe Jackson!

[Nursie]

there was a blern-hitting machine!

Re:Wireless Joe Jackson!

[CustooFintel]

Exactly! He was a machine designed to hit blerns!

the secret is

[martas]

there's actually a really skinny guy hiding in the "robot". but seriously, it's quite impressive -- might not seem like much now, but the thing about robotics is that getting the basics is the hardest part; once you have that, getting fancy is relatively easy.

Re:the secret is

[msobkow]

Back in the early '80s, I remember being impressed by a pogo-stick hopping robot just because it could keep it's balance. That's amazing progress for 30 years when you think about it.

Re:the secret is

[mangu]

Back in the early '80s, I remember being impressed by a pogo-stick hopping robot just because it could keep it's balance. That's amazing progress for 30 years when you think about it.

If there existed any use at all for a robot hopping on a pogo stick they would have developed it further. I remember the robot you mention and I also remember thinking why would anyone want that.

A robot capable of following visually a moving ball is something entirely different. Tracking an object over an arbitrary background would allow some extremely useful capabilities.

May I mention robots-driven vehicles on normal streets?

Re:the secret is

[jklovanc]

Sorry by in my experience it has been the exact opposite. The basics are the easy part but when you try to deal with the last 20% one finds that there are many exceptions to the basics. When the exceptions interact they spawn more exceptions making things even more difficult. It is the old 80/20 rule. You get 80% of the job done quickly but that last 20% can make the 80% unworkable.

This robot can stand in one place and deal with a ball that crosses 2/3 of the baseline in a restricted altitude. Here are some issue;
1. What if the ball would bounce twice before crossing the baseline (drop shot)?
2. What if the ball can not be reached from the stationary location (say it goes off the side of the table or high over the robot's head).
3. What if the ball comes faster than the robot can move(physical limitations)?
4. The game of ping pong is about strategy and anticipation which in robot terms means AI which is much more difficult than rallying.

So in effect the robots can rally under specific parameters and nothing else. Cool but not really impressive

Re:the secret is

[martas]

You're absolutely right, but I guess I was talking about a much larger definition of "the basics". E.g. if you could make a robot that is a mediocre soccer player, or waiter, or pilot, etc, then chances are you've done most of the work required to make an exceptional soccer player, or waiter, or pilot, etc.

Re:the secret is

[cffrost]

If there existed any use at all for a robot hopping on a pogo stick they would have developed it further. I remember the robot you mention and I also remember thinking why would anyone want that.

With appropriate materials and engineering, such a robot could be used to clear land mines, with the benefit that it could continue operating without intervention after detonating a mine and landing pogo-stick down.

Re:the secret is

[Anonymous Coward]

That would have been Marc Raibert at CMU, later at MIT. You might have heard of him, he's head of Boston Dynamics (cf. Big Dog).

The thing to note here is that the pogo robot was an important part of understanding how balance, dynamics, and robotic control work. Later, that knowledge was used at the MIT lab and then at Boston Dynamics. It's part of the entire learning process that is science. Nobody really wants a pogo stick robot (well, it would be cool, but not worth the effort); it's the science that matters. Perhaps we should remember this when we're thinking about what to cut back on.

Terminal man

[MichaelSmith]

Brings to mind a scene in Michael Crichton's book Terminal Man where somebody is developing a ping pong playing robot (IIRC called HAPP (Hopelessly Articlated PingPong Player)) and the point is made that the ability to accurately deflect a table tennis ball could have all kinds of defense applications.

Re:Terminal man

[Raenex]

If wars were fought with ping pong balls, sure. I don't see how it helps against bullets and missiles.

Re:Terminal man

[0100010001010011]

The only difference between ping pong balls and bullets is scale. If you got this thing good enough to pick out and detect bullets it'd be like any Hollywood action movie.

Re:Terminal man

[Raenex]

Scale matters. The technology to detects bullets and deflect them is going to be drastically different than ping pong balls. You're going to need high-speed cameras and a mechanism fast enough to deflect it.

A robot arm that mimics a human one enough to play ping pong is not even close to the right solution. It's so wrong that you're just wasting time by even trying to solve that problem if what you really wanted to do was stop bullets.

Re:Terminal man

[hedwards]

I'd go so far as to say that it's nigh impossible to do that. They design Type IIIA bullet proof vests for use against a projectile moving 427m/s, so in order to give that kind of protection to a robot, you'd have to have an impossibly fast reaction time or for the projectile to be originating from a kilometer away. At which point it becomes questionable that the detection system would even be able to detect and model the motion while still having time to deflect it.

Granted with lower velocity bullets it's somewhat easier, but still, I'm not so sure that you could move a robot enough in that fraction of a second to make any meaningful use of the technology.

Re:Terminal man

[Raenex]

Theoretically it's possible to detect the bullets, since light moves at about 300,000 km/s. If a fight is occurring at about 100 m distance, you have a quarter of a second to respond to it using 400 m/s as the bullet speed.

For deflection, you definitely would not be using anything remotely like the technology used for this ping pong robot. Instead of moving an arm, you'd likely be firing a counter-bullet, but you'd need the mechanism to be extremely fast to aim, some kind of bullet launcher built into the body, or perhaps an array of them.

The real trick would be detecting all these bullets all the time. Current high speed camera setups are expensive and extravagant affairs, as far as I know, and not something you can just stick on a robot and let it run indefinitely, or even for several minutes.

Re:Terminal man

[Anonymous Coward]

Funny you should mention that. A cousin of mine was involved in a 1980's graduate school research project to develop a tennis-playing machine. It was not humanoid in form at all, but looked more like the large, articulated robotic arms used on automobile assembly lines for welding. The device was supposed to be able to swat at tennis balls ... so said the lead researcher-professor. My cousin was shocked and disillusioned into to dropping out of graduate school when he inadvertently became privy to documents indicating that the device was actually intended to swat hand grenades -- he was very committed to peace, you see, and the combination of working on a military project and being deceived about it really got to him.

Re:Terminal man

[Anonymous Coward]

You don't need a high speed camera, you need laser detection. The setup to detect a single bullet would be a lot cheaper than using high speed cameras. With a single laser, you could scan a fairly large area quite quickly, divide an area up into sectors, using several lasers and detectors, and you can very accurately, and very quickly detect the velocity and location of many bullets.

You don't even need very clever heuristics (which you would need with a high speed camera). With a laser, you know the velocity with very high accuracy, so it's very easy to detect a bullet. 0.4s. I you have a 1GIP (billion instruction per second) processer, that is enough time to run 400,000,000 instructions. At a wild guess, to accurately detect a bullet's velocity and location, would take no more than 1000 instructions, deciding what to do (i.e. the angle to put the deflecter at, or where to shoot the counter-bullet) would take less than 100 instructions. So the processing and detection speed is moot, so you're basically constrained only by how long it takes the mechanical parts to move.

Re:Terminal man

[Anonymous Coward]

If wars were fought with ping pong balls, sure. I don't see how it helps against bullets and missiles.

Instead of a paddle, the robot is holding a 9" titanium spike. It views your eyeballs as two ping pong balls. It's not designed to block any bullets or missiles that that you might be firing at it, it's designed to close the 25 foot distance as rapidly as possible and embed the spike in one of your ping pong balls as quickly as it can, or at least that's likely to be one of any number of military perspectives on applications for bipedal robots that are faster on their feet than humans and can play ping pong well.

And

[JustOK]

Has anyone asked the ping pong balls what they thought of all this? Probably not.

gonna need a faster camera

[Spy Handler]

before the robot can compete with human players... 30 fps aint gonna be enough if somebody were to smash the ball hard.

Re:gonna need a faster camera

[Zouden]

That's why these robots have 120fps cameras.

Re:gonna need a faster camera

[garaged]

It is enough for humans, right?

Re:gonna need a faster camera

[Anonymous Coward]

30fps is enough to get video to look to acceptably smooth for humans, but human vision does not have a particular framerate, just like it does not have a specific resolution. Also, I am not sure how the reaction time of the robot compares to that of a human: a faster camera might give it an edge and let it at least get an image of the ball earlier than a human would even if it has to take more time to process that image and decide what to do.

Yikes robot plays better than me

[youn]

I better practice then :)

What does it look like when it fails?

[The New Andy]

I wish the videos showed what happened when it stuffed up. This is impressive, but seeing how far you can push it before it fails would let me know the upper bound of its impressiveness :-).

Re:What does it look like when it fails?

[Ihmhi]

It's a Chinese robot, they aren't allowed to show it failing. The only way you'd see this robot fail is if it was somehow involved in a disaster that killed a hundred people.

Hatsune Miku in TFA

[Kandei]

The second video is a news broadcast; they play a snippet of Ievan Polkka. Gad, I love Chinese piracy.

Re:Hatsune Miku in TFA

[Anonymous Coward]

What exactly makes this to automatically piracy?

Vital statistics

[snipergirl]

"They are about 1.6 meters tall and weigh in at 55 kilos" Holy shit, they really ARE ping pong players!

oblig

[gyepi]

I, for one, welcome our new ping pong overlords!

Re:oblig

[Kyont]

1.6 meters tall... 55 kilos in weight... impassive facial expressions and unfashionable Chinese jackets... playing mediocre table tennis... my god, they've invented San Jose State!

Controller

[vegardh]

Is the human in the middle controlling something with that thing he's holding?

Exciting

[vagabond_gr]

Now you only need to fill the arena with a bunch of robot-fans programmed to act excited.

Pong

[Zoxed]

Makes me feel old: when I was a kid the best computers could do was play Pong against me !!

Re:Pong

[camperdave]

Makes me feel old: when I was a kid the best computers could do was play Pong against me !!

... and still beat me.

Another channel of feedback: Sound.

[bubulubugoth]

Do they only manage optical tracking? If you ask any sport practitioner, not only the vision is used, also sound feedback is very important, you can determine the strength, even the effect with a given sound...

I wonder if they are including it already...

Yoshimi battles the ping pong robots

[Anonymous Coward]

Yoshimi is a black belt in karate. I know she can beat them!

Special markings on the table

[DavMz]

If you look at the video, you can notice special markings on the robot's side of the table: there are 2 rows of 4 dots parallel to the center line, which I guess are here to help the robots to estimate the position / trajectory of the ball. But this also make the table not conform to regulations (see specifications of tables for table tennis).

Re:Special markings on the table

[jklovanc]

I saw those too but then noticed that they were not on both sides of the table. I bet that they were put there during early development and testing of a single robot. The robot was on the side of the table with the marks and the human serving, who didn't need the marks, was on the other side . When development reached a certain point the marks were no longer needed but they neglected to remove them.

How does the visual tracking work?

[compwizrd]

If i wear a shirt the same colour as the ball and get too close to the robot does it start smacking me instead?

Re:How does the visual tracking work?

[Osgeld]

it should

Lack of historical knowledge

[johnwbyrd]

Yet another failure of Slashdotters to research a topic before posting.

Russell L. Anderson made robots play ping pong at AT&T Bell Labs in 1988:

http://www.ieeecss.org/CSM/library/1989/feb1989/w15-21.pdf

Re:Lack of historical knowledge

[Anonymous Coward]

Yeah, but Anderson didn't upload it to youtube...

Re:Lack of historical knowledge

[Anonymous Coward]

Somebody mod the parent way way up

Not even Skynet ping-pong robots...

[jensend]

have anything on Bruce Lee playing table tennis with nunchaku.

Seeing that movie changed my life ambitions. Fo' rizzle.