Wednesday, February 15, 2012

Y&T S-27 table tennis robot

You go back, Jack,
do it again...
wheels turning
'round and 'round
You go back, Jack,
do it again...

I have a Y&T S-27 programmable table tennis robot.

It is a high-end robot made in China. In 2010/2011, it was considered state of the art. I bought it in Zhongshan (中山), China back in 2010 and had it shipped home.

I only take it out about once a year because I prefer to play with people. Human beings have more natural variation, which can be both a blessing (helps train adaption) and a curse (inconsistency). The other big problem with the robot is that it doesn't have arms: you cannot see it swing at the ball like a human being. A large part of anticipation is seeing how the opponent reacts to and contacts the ball. With the robot, the balls just shoot out almost without warning from a tube.

Last night, for the first time in over a year I used the robot to do a drill. (Usually, I'm fortunate enough to be able to find players to come out and hit with me.)

I programmed a drill asking the robot to send alternately one topspin ball slightly to the right of the centerline (position 7) and one to the forehand corner (position 10). One from each physical robot head. I asked it to send 40 balls and stop. (As the video belows shows it actually sent 41 balls.)

The way these robots work is having a spinning motorized roller shoot balls from a tube. The more sophisticated robots use two rollers. The spin and speed of the ball is determined by two (orange colored) spinning motorized rollers: one that brushes the top of the ball and one that brushes the bottom. You get topspin if the upper roller spins faster than the lower one, and underspin if the lower one is faster.

I set the top motor to speed 6, and the bottom to 1. This means the balls comes to me already loaded with topspin. The higher the speeds, the faster the balls.

The frequency of ball delivery was set at the 2nd slowest setting. This produced balls at the rate of 1 per 1.75 seconds. That gives me plenty of time to recover between balls, and perhaps approach the ball a bit more deliberately instead of frantically trying to keep up and reacting like a headless chicken. And as I get used to the drill and learn to anticipate the ball, cognitively I'll be less overwhelmed, and be able to work on improving my form. At least, that's the theory...

The control panel settings are shown in the picture below (left side for left head, right side for right head):

I'm currently trying to retool my forehand loop stroke. In the following video, I'm practicing my counterloop; that is, I return the incoming topspin ball with topspin. Fight fire with fire, so to speak. I'm only partially there. I'm trying to backswing less these days. This saves precious tenths of a second. And I'm think I'm finally succeeding in this aspect.

(There are other strokes possible. For example one could simply block the incoming topspin. Especially if one has less time. See my earlier post and video with Crystal Huang here.)

Also, having the robot send balls alternately to different positions helps me train my feet to respond properly. I have to move. And keep my balance. Loop. And then not have my loop stroke throw off my balance because I need to respond to the next ball. Multiple small hops are vastly preferable to a single larger step because this allows for fine-grained adjustment. You'll be surprised how many hops are possible in half a second.

Maintaining balance conflicts with how strong you can counterloop. The momentary stop for balance on the right leg, and the waist twist and weight transfer are crucial aspects of this part of the counterloop equation. Although my loop percentage rate is satisfactory, judging by the video clip, I could improve quite a bit on the quality of my loop.

Link: here

Moreover, my followthrough is not where I want it to be at the moment. This affects the amount of power delivered tremendously. It's way too much to the right, crossing the imaginary centerline of my body. Where the stroke finishes also affects recovery time. This is a stroke error. Instead I'd like it to finish with the paddle more out to the front. Technically, I also need to work on holding onto the extended stroke position momentarily at the end of the stroke. I automatically drop the paddle at the end of stroke which prevents proper transitioning. This is another error.

These are all good reasons why you want to resist the temptation of setting the machine to shoot as many balls per second as you can cope with. It makes good sense to debug the stroke fully before increasing the tempo.

In summary, the robot is not just useful when you can't find anyone to play with, it's also a good tool for reprogramming muscle memory. The problem of course is making sure the new program being written is a true fundamental improvement over the old. That's where video analysis is extremely useful.

No comments:

Post a Comment