|
|
|
|
Red.13
Joined: 17 Jun 2013
Posts: 47
|
Firstly, thanks to Marto for directing me to this thread. As it is 6 pages long, it will take me a while to get through it all, so please bear with me while I go through it... Great thread... And glad the discussion is being had. 
But, within the first page, I can see a tendency towards complexity, and that is simply not required with a well-designed autonomous robot. (AR for short.)
The hardest part about making an AR is knowing where the enemy is. Emulating "sight" as is perceived by a human is damn difficult if not impossible for something that is to be untethered and self-processing.
So for an AR to ever be able to competitively go up against an RC robot it must be able to "see" the competitor quickly, and preferably identify which is the "front" of the enemy, particularly if they have a drum, saw or other debilitating device on a particular surface. (I will ignore spinners at this stage since I find them uninspiring, boring and frankly unsportsman-like.) Knowing where the enemy is, which way they are moving, and which end is their "front" is 90% of the battle.
To this end, the best way that I have had experience with (to date) is with Infrared emitters, typically cycling some frequency. These are easy to identify on the move.
My initial idea (before reading the other 5 pages ) would be that RC competitors would need to be fitted with an IR emitter on them. Not going to suggest how this is done at this point, other than to say it needs to be able to be "seen" by the AR, pretty much at all times. So maybe a small black dome on top, with 4 IR LEds flashing away. Or the other thought (from the hip) is that the RC needs to have some highly reflective stickers attached to it, so the the AR shines out IR light and can pick up the reflection.
Just my initial thoughts and as I read through the thread, will happily add more if feedback is constructive.
I would hope that one day, we could have combats that have a mix of both AR & RC and the spectators would not know the difference. But to get there, we have to understand the limitations of on-board processing and to be "spartsman-like" to the AR, we might need to assist in small ways to allow them to "see" their competitor.
Mark
|
Sun Jun 23, 2013 1:38 pm |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Red.13
Joined: 17 Jun 2013
Posts: 47
|
quote:
Originally posted by RogueTwoRobots:
quote:
Originally posted by Jaemus:
Autobots? 
A Prime suggestion
3 letters max... Or I will just stick to AR.
quote:
Originally posted by RogueTwoRobots:
I'm not that into autonomous robot side of things, though I have no problems with them. But I would be against putting things on my robot so that an Autobot can 'see' and attack me. I'm in there to win and I don't feel compelled to assist my opponent in trying to defeat me. That's part of the challenge of making one.
I don't get the difference: When you are up against an RC controlled by a human, do you ask them to be blindfolded? (Well, maybe you do, but I doubt many would agree - except the spinners.  )
Right now, "seeing" the opponent is the hardest part for an AR to get over. If this group, or community as a whole, want AR to be competitive, then you will need to let them "see" you. At least in the short term. Later, when processors and sensors get better and faster, then this requirement would not be needed. And until AR's start beating you, do you have anything to worry about?
Mark
|
|
Mon Jun 24, 2013 10:50 am |
|
|
|
|
|
|
|
|
|
|
Red.13
Joined: 17 Jun 2013
Posts: 47
|
quote:
Originally posted by marto:
So why have I not bothered?
...Yes it is more complex...
Really it doesn't interest me that much. Feel free to finish it off.
It is because it is so complex that your interest has waned. And you will notice that with your system (as far as I have seen so far) you need to have some pretty specific markings on your enemies, that some are already objecting to and in some instances, due to the build of the robot, might not be possible, like on a spinner.
quote:
Originally posted by marto:
but anything onboard is really useless without a PC or expensive sensors.
Not at all! A good IR-based distance sensor is only $15. An ultrasonic one costs $4. IR Lasers are a couple of dollars and IR LEDs a couple of cents. Or maybe you need to bathe the whole arena in IR light to reflect off the tape better. (Again, just thinking aloud.)
quote:
Originally posted by marto:
It also means that if you develop such a system release its specs and software and make it easily to replicate you can just install that in your arena and then everyone can use it.
That can still be done... Just like the ArduFlyer system does for copters.
quote:
Originally posted by marto:
I am more interested in finishing something like this off to get automated docking and charging working
Well it so happens that is something that I will be working on as well, and I will be happy to share my results and methods. But it wont have any external systems to help. It will be solely onboard and I am sure it will be possible.
Can you expand on how you want the robot to behave? As in, could the "start position" be anywhere in the ring? In any orientation? And is the end position in a particular corner? In a particular orientation? Is the charging done by contacting something on the wall, or is it done by induction off the floor?
Can the combat ring have some specific markings on it? Like say the floor is all black, but has a white line on the outside of it? (See where I am going? Why look for a complex answer when a simple one is in your face. )
quote:
Originally posted by marto:
Lol this implies that you are actually going to build something. May I suggest you start with an RC box first, so you can try and get past the wheel falling off stage. Just because its RC doesn't mean its easy to do.
Oh, I never said doing an RC is easy to do. But that would most certainly be the way to do it. Putting in the automation is "simply" replacing the human "vision" of seeing the enemy, and then the micro-computer directs the movement of the motors in exactly the same way as you use your sticks. No difference.
But like you, my time is limited, and doing BOTH the building of the robot AND developing a system of "sight" for the robots is taxing. So, why not step back to the old "building blocks" option...
As in... Does anyone have an old RC attack robot they are no longer using they wish to sell?
|
|
Mon Jun 24, 2013 10:35 pm |
|
|
|
|
|
RoboWars Australia Forum Index
-> General Chatter
