We're excited to announce a game-changing feature that's currently in testing for the PongFox table tennis robot. Once rolled out, this innovation will allow players to interact with the ball after service, a capability not yet found in any other commercially available table tennis robots. The crucial aspects of service practice and preparation for third-ball play, integral to table tennis training, are the focus of this enhancement
Advancing Table Tennis Robots: PongFox Tests Service & Return Feature - Thoughts?
- Thread starter palguay
- Start date
This user has no status.
This user has no status.
Member
Does one of the amicus robots have a feature when it hears you serve, it can return it? I'm pretty sure I heard that somewhere.
This user has no status.
Active Member
This user has no status.
This user has no status.
Well-Known Member
I second that. No. I have Power Pong which is basically amicus robot (they both were built in the same factory). You can set up drills but it cannot hear you serve.
Another advantage with the pad is that we can now add a feature to track you consistency which can be used as an assessment for your progress over time
This user has no status.
This user has no status.
Well-Known Member
Good idea, but the 1st feed versus a realistic serve return timing is going to be very difficult to get right.
if the feed timing is too slow, then come game play, you will be late for your 3rd ball. (so will this feature make you better, or make you worse?)
what happens when you serve short underspin and the opponent drops the ball back, There is timing, and strength of ball placement to think about. You reduce the speed coming to you, the ball timing will be even further off.
Sensor over the net would be a better idea than that of a small pad on a big table.
if the feed timing is too slow, then come game play, you will be late for your 3rd ball. (so will this feature make you better, or make you worse?)
what happens when you serve short underspin and the opponent drops the ball back, There is timing, and strength of ball placement to think about. You reduce the speed coming to you, the ball timing will be even further off.
Sensor over the net would be a better idea than that of a small pad on a big table.
We have earlier created videos on calculating the speed of rubbers, we use a similar technique to get the numbers and fine tune it. also the robot can be moved around the table as well as each strokes can be programmed to program the type of return required and options of choosing different returns to get some interesting options of training.
today when practicing service most people just keep a bunch of balls and serve , this pad forces two things one is to get the position of your serve right and the second is to work on transitions changing grip back in some cases and getting ready and playing.
Having a pad also gives an option to train on consistency when using without service and ability for a player to track it over time
I think over time smart coaches will look at such tools and figure out in which cases it works well and where it wont and use these technologies to make their jobs easier rather than thinking about it as a replacement for real training.
This user has no status.
This user has no status.
Well-Known Member
So if i want to serve long, i push the robot back to where it is like in your video.
I want to serve short, do I pull it closer towards the net?
1 short, 1 long, do I then maybe put the robot on some rails?
Respect your thought and opinion, here is mine.
I think coaches without ability or capacity will make use of additional gadgets like this.
or places that have unformal training.
Any formal training facility or even amateur clubs, you will just have 2 practice partner, one practicing serves and the other serve return and then open play.
To me, there isn't anything too "smart" I see from the video. Robot is for 1 man shows, not for teams.
and because of that, proper training centres not using it, doesn't mean those coaches are not smart. The ones using it, doesn't equate to them being smart too.
If numbers make such a big differences, it will be stats during match play and not from your 20 or 30 hours training a week. Or like me doing 7 hours of serve training a week at my peek training schedule (that will be too much data for me to waste my time reading, while I should be resting and relaxing my brain)
Visual sensor should be the new wow factor, and ball to be fed the moment the ball crosses the net, not like it is now, landing on a pad and then wait seconds longer than a realistic return.
Especially, if you serve long like that, the returns will come a lot faster and at an angle too from where you served the ball to, not from the center of the table.
These are facts and a smart businessman should take it in and better the product.
In table tennis, Humans still is better than Robot. Can't deny that.
But if you are a 1 man show in your garage.... something is better than nothing, but if the feed is too delayed, then it ain't helping you.
This user has no status.
This user has no status.
Well-Known Member
any comments from the respectable coaches here?
a pad that senses the ball land and then triggers a "serve return" feed.
I gave my feedback to the owner on another thread (hope you can answer without seeing my feedback to get a neutral opinion), I know many coaches only post in this thread, hence why I am posting and asking.
Thanks a mill
This user has no status.
Well-Known Member
Oh I can see a lot of game reading landmines. The ball coming from a location where the serve did not go, the spin and speed on the returned ball not bring determined by the serve, no visual cue to read spin on the turn etc.
But maybe there are some benefits. I find that people csn improve with these things in ways I can't imagine but I think the negatives don't make it a reasonable substitute for real third ball drills.
This user has no status.
This user has no status.
Well-Known Member
The robot head can be moved behind the pad. The pad tries to solve a problem I have with robots and that is timing. What would be more interesting is having an accelerometer mounted to a catch net so when the ball hits the net, another ball is shot. All the accelerometer needs to do is detect vibration from the ball hitting the net. This way the net and robot could be moved back from the table to simulate an opponent that is hitting from behind the table. One should be able to return the ball so that most go back into the net even if the net is back 10 ft/ 3m
This user has no status.
This user has no status.
Well-Known Member
I would love that the robot can move spots, so you can do 1 long and 1 short serve. I guess, you are fixed to 1 spot only and then the manual labor moving the robot around....
The net+accelerometer is a good idea, but if one does very powerful top spin shots, the ball will go into the net and spin in circles - meaning the moment the second shot is hit again into the net, the 1st ball is still moving and could still be causing vibration.
says
2023 Certified Organ Donor
says
2023 Certified Organ Donor
Well-Known Member
This user has no status.
Well-Known Member
Depends on the playing level. I could argue that a basic robot has value for training all levels of movement. But this tool could cause serious problems for a player higher rated than even US 1500 IMHO. I would think even US 1000, but it depends I guess.
says
Spin and more spin.
says
Spin and more spin.
Well-Known Member
Moderator
I moved this subject to its own thread because it is not a subject about one specific person's footage and to restrict comments from people who are not on the authorized list would not be fair, especially when the comments are useful.
NextLevel makes excellent points. So, I would agree with his comment. Also, a coach can read what a player needs to work on and provide that. I don't think AI is going to be able to know a player's weaknesses and provide a constructive progression to help the player improve for some time in a sport like TT.
NextLevel makes excellent points. So, I would agree with his comment. Also, a coach can read what a player needs to work on and provide that. I don't think AI is going to be able to know a player's weaknesses and provide a constructive progression to help the player improve for some time in a sport like TT.
This user has no status.
Well-Known Member
I know someone working on the problem and let me just say that for stroke technique, you might be surprised.
says
Spin and more spin.
says
Spin and more spin.
Well-Known Member
Moderator
Stroke technique, sure.
But the things that happen in random play where a player gets put out of position and has a habit that limits their ability to create certain shots....seeing the flaws that are brought out in game strategy, not sure there is a robot that can read that stuff and help a player fix it yet.
But the biggest issue is, seeing the opponent and what he does with the racket and how he creates shots, you don't get to read that stuff from a robot.
They are good for some things as people have pointed out. I definitely used a robot as part of what helped fix a broken FH stroke. But it did not help me read the opponent and it did not help me turn open to my FH better when I was locked in on my BH in a rally with a person.
This user has no status.
This user has no status.
Well-Known Member
Yes, but it is easy add a timer so the rattling around of the incoming ball is ignored for a period of time. In industrial controls we call this "debouncing" ... REALY! This is a common problem in many controls situations.
An accelerometer could be mounted to the table. It could detect when the ball bounces easily. Determining where the ball bounces on the table is much more difficult. Also, the computer would need to be able to tell if the bounce is the first bonce of a serve. This isn't hard because there is a long time interval before the next serve.
Determining were the ball bounces or where the ball is gets to be very expensive because it involves vision systems.
This user has no status.
Member
This is an interesting idea. Reminds me of what I saw from Tom Lodziak's video with robot training.
Here it is:
<- Regarding the video I posted, I don't know how I feel about it being so thick. I think having a lower profile sensor would be the way to go if you can get it into that form factor.
Tony's feedback of getting match data (ball placement, even human positioning) would be pretty valuable as a value add for the next step in your research/production.
Here it is:
Tony's feedback of getting match data (ball placement, even human positioning) would be pretty valuable as a value add for the next step in your research/production.
This user has no status.
Member
Since TT is filled with conditions, it would be interesting to create a Bayesian model to analyze a person's serve and predict the likelihood of what the server will give. Something like this paper for tennis, but applied to table tennis would be interesting: https://summit.sfu.ca/item/34899
From there, it would get rather complicated because you're coming up with a lot of situational context.
Granted, the system would be entirely different that what was proposed. Would need more of a grid sensor system or a computer vision system to get that going, so that just adds on to the expense. Even when practicing something like a Backspin Pendulum, it would be helpful to understand where the placement ends up going when drilling. The output can even be represented as a heatmap.
From there, it would get rather complicated because you're coming up with a lot of situational context.
Granted, the system would be entirely different that what was proposed. Would need more of a grid sensor system or a computer vision system to get that going, so that just adds on to the expense. Even when practicing something like a Backspin Pendulum, it would be helpful to understand where the placement ends up going when drilling. The output can even be represented as a heatmap.
