Following on a well-publicized match between Timo Boll and a Kuka robot arm that was, being kind, a videographer's realization of what it might be like, it seems that doubt has fallen on all ping pong playing robots. Is this fast and smooth creation real or fake?
First take a look at The Duel - the video that started it all:
Yes, you can see that this is more an ad for the robot than any proof of the ability of the robot to play anything in autonomous mode. The camera angles and cuts never let you see a rally, just individual shots that could be, and probably were, staged.
A nice idea but ultimately a disappointment.
Now let's take a look at a much lighter and smaller, but much more impressive, robot playing ping pong:
It really is impressive but is it real or CGI?
Evan Ackerman in the IEEE Spectrum column "Video Friday", a great place to find robot videos every week, thinks it is a fake. In fact he refers to it as:
" fake fake fake fake fake"
Which is very definite.
He provides the following justification for this conclusion:
"Besides the fact that I'm highly skeptical that this is the kind of high-speed motion tracking that you can do in your garage with tiny little cameras, you'll notice two things about this video: first, despite the mass of the robot zipping back and forth and changing direction quite quickly, the table doesn't shake in the slightest, and the net doesn't even vibrate. But the real giveaway is right at the beginning: if you pause it in the right places (there are several, one is right before 0:06), you'll see pieces of the robot (usually its elbow) disappear and reappear. Yup, not real."
The speed is amazing but not impossible and the same is true of the motion tracking. A white ball against a green surface makes it easy. The math needed to position the arm is something else again and enough to give anyone a headache, but again you can see that it is do-able. Watching it carefully the dynamics and prediction seems to be right - the robot arm never seems to show pre-cognition by moving to intercept a ball before it is hit, for example. The consensus here at I Programmer is that it is difficult but possible.
The final piece of evidence is more difficult to weigh up. Something happens that is odd at about 0:06 to the image, but what exactly is difficult to say.
If you go to the blog (in German) you will find a fairly convincing account of how the robot was built. There aren't enough instructions to know how it works in sufficient detail to reproduce it, but it seems convincing.
So is this real?
We all hope it is because it demonstrates what is possible today on a budget and some skill.