Be nice to know what the fail criteria are...I reckon any of the red areas touching the floor are, and any o fthe connecting beams too, but it also seems to fail if its not in contact with the floor for a certain period of time...
Submitted by babychaos on Tue, 2008-12-09 13:27
The writing bloke said "This is a GA I wrote to design a little car for a specific terrain. It runs in real-time in Flash. The fitness function is the distance travelled before the red circles hit the ground, or time runs out. The degrees of freedom are the size and initial positions of the four circles, and length, spring constant and damping of the eight springs. The graph shows the 'mean' and 'best' fitness"
Submitted by Evilmatt on Tue, 2008-12-09 13:48
Ahhh..the time limit would explain it...all my buggies were going off a big jump when time ran out I guess...
Submitted by babychaos on Tue, 2008-12-09 13:54
Been running mine for about 4 hours now and it's slowly creeping up. Looks first order to me.
Submitted by brainwipe on Tue, 2008-12-09 14:08
It's weirdly hypnotic...
Submitted by Nibbles on Tue, 2008-12-09 14:30
mines become self aware and has escaped the boundries of the flash window and doing burnouts and doughnuts all round my screen
Submitted by Evilmatt on Tue, 2008-12-09 14:44
What you really want it some kind of way of comparing your own evolved buggy. With the small number of variables it would be fairly easy to codify it, and then add it in as a parameter to a 2-player version...
Submitted by babychaos on Tue, 2008-12-09 14:56
I can see a webgame based off this you get a random buggy that you can enter into races against other buggies people that win get the offspring of their buggy and the other racer :D the loser can either keep their current buggy or get a new one.
Interesting mine looks a fair bit different to yours rob
it is longer and has a triangular structure with no cross links so the frame flexes as it drives along
Submitted by Evilmatt on Tue, 2008-12-09 15:44
I didn't get a shot of mine, but it was again different, with a trapeziodal frame, smaller at the top.
I'll run it overnight tonight, and see what it comes up with...
Submitted by babychaos on Tue, 2008-12-09 15:50
I've got three new copies of it running in tabs and at least so far it seems to be taking a different approach in each.
Submitted by Evilmatt on Tue, 2008-12-09 16:10
I did the same as Matt this morning at 8am, and kicked off 3 simulations. As of 1.30pm this morning here is how the 3 are getting on...
One has developed the cunning ploy of throwing itself over a sharp point in the cliff, and throwing it's backend far forward just within the time limit...how thats managed to stay it's best tactic over 5 hours is anyone's guess.
The second is a fairly rigid machine that struggles over bumps. This ones seems to change fairly regularly, but hasn't yet hit a decent design.
The 3rd has a massive front wheel and is very bouncy, but seems to get a lot of failures within each generation, where the red balls fold through and hit the floor. The weight over the front wheel keeps the grip down, so this one is getting furthest so far.
He had anote up yesterday saying he would release the code at some point...if he does I'll happily take it and slap some output ability on it so that vehicles can be raced. It would be good to see how he did the hit detection as well...Flash didn't deal with that very well when I used to use it...
Submitted by babychaos on Wed, 2008-12-10 13:59
Mine computer unfortunately crashed and I lost the progress but I had gotten two similar to your 2nd and 3rd. The other two were one that was front loaded with a big front wheel with the weights over it and a little back wheel with an equilateral triangle shape. And the other was two large wheels very close together with the weights sandwiched in the middle.
Comments
Thats rather good...
Be nice to know what the fail criteria are...I reckon any of the red areas touching the floor are, and any o fthe connecting beams too, but it also seems to fail if its not in contact with the floor for a certain period of time...
The writing bloke said "This is a GA I wrote to design a little car for a specific terrain. It runs in real-time in Flash. The fitness function is the distance travelled before the red circles hit the ground, or time runs out. The degrees of freedom are the size and initial positions of the four circles, and length, spring constant and damping of the eight springs. The graph shows the 'mean' and 'best' fitness"
Ahhh..the time limit would explain it...all my buggies were going off a big jump when time ran out I guess...
Been running mine for about 4 hours now and it's slowly creeping up. Looks first order to me.
It's weirdly hypnotic...
mines become self aware and has escaped the boundries of the flash window and doing burnouts and doughnuts all round my screen
What you really want it some kind of way of comparing your own evolved buggy. With the small number of variables it would be fairly easy to codify it, and then add it in as a parameter to a 2-player version...
I can see a webgame based off this you get a random buggy that you can enter into races against other buggies people that win get the offspring of their buggy and the other racer :D the loser can either keep their current buggy or get a new one.
Here's mine after about 5 hours 20 minutes ish.
I could watch this for ages...
www.aserash.com
Interesting mine looks a fair bit different to yours rob
it is longer and has a triangular structure with no cross links so the frame flexes as it drives along
I didn't get a shot of mine, but it was again different, with a trapeziodal frame, smaller at the top.
I'll run it overnight tonight, and see what it comes up with...
I've got three new copies of it running in tabs and at least so far it seems to be taking a different approach in each.
I did the same as Matt this morning at 8am, and kicked off 3 simulations. As of 1.30pm this morning here is how the 3 are getting on...
One has developed the cunning ploy of throwing itself over a sharp point in the cliff, and throwing it's backend far forward just within the time limit...how thats managed to stay it's best tactic over 5 hours is anyone's guess.
The second is a fairly rigid machine that struggles over bumps. This ones seems to change fairly regularly, but hasn't yet hit a decent design.
The 3rd has a massive front wheel and is very bouncy, but seems to get a lot of failures within each generation, where the red balls fold through and hit the floor. The weight over the front wheel keeps the grip down, so this one is getting furthest so far.
He had anote up yesterday saying he would release the code at some point...if he does I'll happily take it and slap some output ability on it so that vehicles can be raced. It would be good to see how he did the hit detection as well...Flash didn't deal with that very well when I used to use it...
Mine computer unfortunately crashed and I lost the progress but I had gotten two similar to your 2nd and 3rd. The other two were one that was front loaded with a big front wheel with the weights over it and a little back wheel with an equilateral triangle shape. And the other was two large wheels very close together with the weights sandwiched in the middle.