I for one welcome our new robot overlords.

>> ^Ti_Moth:

I for one welcome our new robot overlords.


I like balls.

That's a weird definition of juggling.

Question: why isn't it perfect?

Is there a "chaos" factor that throws off otherwise exact calculations?

Or are the calculations "dumbed down" for speed?

>> ^COriolanus:

Question: why isn't it perfect?
Is there a "chaos" factor that throws off otherwise exact calculations?
Or are the calculations "dumbed down" for speed?


Air movement and imperfections in the balls an paddle.

It does seem to find a sweet spot, which is why the 3 balls eventually collide.

Well it says the paddle stays flat and never tilts. So that means that in order to change the direction of the ball, the paddle must be moving at the point the ball hits the paddle in order to apply spin to the ball and get it to change direction. That probably makes this harder to engineer than it initially appears.>> ^COriolanus:

Question: why isn't it perfect?
Is there a "chaos" factor that throws off otherwise exact calculations?
Or are the calculations "dumbed down" for speed?

Washington? Optimal Control? Must be Todorov.

When you read the motor control literature, you quickly realize the importance of uncertainties. They are pervasive. Every sensor has a resolution, and every motor command given has a latency. That means that you are always using imperfect data to predict some future state of the world. The further ahead you try to predict, the more small errors in measurement become large errors in prediction. So no, these aren't exact calculations. Like most motor control problems, it is probabilistic. In fact, optimal motor control in this context (Todorov's primary line of work) generally refers to a control scheme that is optimized to minimize endpoint error. The fact that there is error to minimize implies that these calculations aren't exact.