I'm using Boxing as the example Combat Sport since it doesn't use kicks and grappling:
Originally Posted by W. Rabbit
An averige boxing round is 5 minutes.
A decent Boxer can throw about 5 punches in a second.
Offcourse this would be the maximum peak performance of 5 punches/second and this would not be done for the constant 5 minutes.
But I have seen Boxers hit that peak on an averige of 3 times in a minute for a round of 5 minutes making that your system would have 15 processing peaks with a total refresh rate of 0.2 seconds.
Even if you would make an interval of 19 seconds between the peaks (which would make your drone predictable) and limited the sensory input...the sheer data that would garantee that the drone would move correctly would be too much for any 'PC' system to process.
There is a reason why Multinationals (especially Banks) stay with Mainframe and Midframe systems: because of their capabilty of processing huge amounts of data a second.
Originally Posted by Jiujitsu77
Originally Posted by Humanzee
Originally Posted by jk55299 on Keysi Fighting Method
The real deadly:
That's the one. When I saw it, an engineer was still putting it all together. They were gonna do a little demonstration later that day but it was hours later and I didn't have the time.
Originally Posted by W. Rabbit
Ok thanks for the numbers.
Originally Posted by Zendokan
This is definitely not a PC project, I am looking to use an embedded system with ASICs and controller area networks that could easily dwarf the processing power of a single PC. Definitely not using Windows :XXbunny:
Last edited by W. Rabbit; 11/12/2010 7:51pm at .
Yeah...too bad...I don't think incendiaries will be practical for sparbot™
Originally Posted by Lord Skeletor
Spent the day looking at robotics designs from Japan, the US, Germany, and Great Britain (amongst other distractions ...:blob8:).
Now am thinking that a great way to get around the "predictability" of the Type 1 (static punching bot) is to just add limbs.
I got this idea after review of designs, during which my 3 year old came in claiming to be General Greivous from "The Clone Wars" and wielding two toy lightsabers.
Had an aha moment.
A four (or even three)-limbed beasty might make a far more interesting/useful standing bot. Why limit it to 2 arms?
The four limbs could each be in a defensive (i.e. "Hit Me!") mode, each essentially in punching pad "target" mode, with one limb randomly firing at an "appropriate time" to approximate counter-punching. Whether or not a limb "fires" could be tuned based on skill level, with beginner modes maybe throwing a counter punch every 10 human strikes on avg (or whatever) to high expertise settings such as the bot attempting to throw more punches than the user.
Without worrying about machine vision this seems incredibly basic to do but would make a really neat sparring machine. Combined with basic machine vision and "follow" logic, the bot could easily be made to move in 360 degrees to continually try to maintain a bearing on its target.
So far none of these approaches would really have any processing bottlenecks, especially for an embedded system.
Did some quickie Phun 2D (physics simulation) runs. Pretty easy to make rudimentary limb units that would approximate the ingress vector of punches in a variety of styles.
These are the major punches I would likely model arms for, unless someone sees something important missing...let me know.
Right now I don't see a great way to post these Phun sims but I can post screenshots after I get more done, or turn a sim into a rotating GIF if I'm not too lazy.
With out machine vision how exactly is this thing aiming its punches?
The idea is that the device is part of a larger "arena". A good illustration is how the Dance revolution machines have a player field that synchronizes with the overall system, and the user/computer interact via two distinct spaces: the control board on the floor and the monitor/audio. Likewise without machine vision (which is not too difficult to have in a project like this, out-of-the-box following robot kits are pretty cheap), there would be a certain "static" limitation to the fighter, but no more than is currently seen when using punching gym equipment like heavy bags or speed bags.
Originally Posted by goodlun
A Type 1 without machine vision would essentially involve a person entering a "box" within a certain distance/bearing of the bot and would perform the same types of exercises they might do with a speed bag or punching bag or other types of equipment, but the bot adds the element of interaction: targets pads can move on their own, there would be counter punches for the player to avoid, you could even add a point-based scoring system that adds points for clean target hits but deducts them for any time the bot registers a good strike, etc.
Type 1 could be envisioned as more like a game system with a robot opponent. The new motion-based home entertainment systems are trying to do the same thing with motion capture....but that type of thing is not what BS type folks really want, I'd wager.
Adding machine vision would effectively create a 360 player arena in ring shape around the stationary bot.
Taking that idea one step further to mobile/wheeled bot (we are getting pie in the sky again but that's cool with me), you'd increase that "ring arena" to a full room, and you could have a bot capable of 360 degrees of motion that could conceivably move with a human target. They already do this with current Battlebots, but those are all tiny, floor roving models. If you made a standing version for people, you'd step into a room, switch this thing on, and it would start working around the room, keeping its facing to you, backing up as you punched targets, advancing when it counterpunches, and so forth.
Last edited by W. Rabbit; 11/15/2010 5:24pm at .