Gaming Steve Message Board

Will Wright's Spore => Spore: General => Topic started by: Daxx on January 03, 2008, 01:41:36 pm

Title: NERO
Post by: Daxx on January 03, 2008, 01:41:36 pm
Now, bear with me for one moment. This thread might be better suited to the PC Games section, but I thought it would gather more attention in here whilst we're waiting for release date announcements and so forth.

Neural Evolving Robot Operatives (or NERO (http://nerogame.org)) is a machine learning game where you evolve your own team of robots and train them up to fight. The robots have neural nets which when training are affected by a form of artificial selection. Since you influence which bots survive and which do not, you can evolve a team on your whim, training them to perform various tasks in battle.

Quote from: NERO website
NERO is an example of a new genre of games, called Machine Learning Games. Although it resembles some RTS games, there are three important differences: (1) in NERO the agents are embedded in a 3D physics simulation, (2) the agents are trainable, and (3) the game consists of two distinct phases of play. In the first phase individual players deploy agents, i.e. simulated robots, in a "sandbox" and train them to the desired tactical doctrine. Once a collection of robots has been trained, a second phase of play (either battle or territory mode) allows players to pit their robots in a battle against robots trained by some other player, to see how effective their training was. The training phase is the most innovative aspect of game play in NERO, and is also the most interesting from the perspective of AI research.

The artificial neural network "brains" learn by means of the NEAT or Neuro-Evolution of Augmenting Topologies algorithm. Neuroevolution is a genetic algorithm, i.e. a reinforcement learning method that operates by rewarding the agents in a population that perform the best and punishing those that perform the worst. In NERO the rewards and punishments are specified by the players, by means of slider controls. The genetic algorithm decides which robots' "brains" are the most and least fit on the basis of the robots' behavior and the current settings of the sliders.

More information can be found here (http://z.cs.utexas.edu/users/nn/nero/about.php), and on the NEAT Wikipedia Page (http://en.wikipedia.org/wiki/NeuroEvolution_of_Augmented_Topologies)

NERO 2.0 has been released fairly recently and can be found from the game's website here (http://z.cs.utexas.edu/users/nn/nero/download.php). I think it would be quite interesting to see what GamingSteve forum members can come up with, and maybe have some battles amongst ourselves!
Title: Re: NERO
Post by: Ultramarine on January 03, 2008, 01:49:18 pm
Now, bear with me for one moment. This thread might be better suited to the PC Games section, but I thought it would gather more attention in here whilst we're waiting for release date announcements and so forth.

Neural Evolving Robot Operatives (or NERO (http://nerogame.org)) is a machine learning game where you evolve your own team of robots and train them up to fight. The robots have neural nets which when training are affected by a form of artificial selection. Since you influence which bots survive and which do not, you can evolve a team on your whim, training them to perform various tasks in battle.

Quote from: NERO website
NERO is an example of a new genre of games, called Machine Learning Games. Although it resembles some RTS games, there are three important differences: (1) in NERO the agents are embedded in a 3D physics simulation, (2) the agents are trainable, and (3) the game consists of two distinct phases of play. In the first phase individual players deploy agents, i.e. simulated robots, in a "sandbox" and train them to the desired tactical doctrine. Once a collection of robots has been trained, a second phase of play (either battle or territory mode) allows players to pit their robots in a battle against robots trained by some other player, to see how effective their training was. The training phase is the most innovative aspect of game play in NERO, and is also the most interesting from the perspective of AI research.

The artificial neural network "brains" learn by means of the NEAT or Neuro-Evolution of Augmenting Topologies algorithm. Neuroevolution is a genetic algorithm, i.e. a reinforcement learning method that operates by rewarding the agents in a population that perform the best and punishing those that perform the worst. In NERO the rewards and punishments are specified by the players, by means of slider controls. The genetic algorithm decides which robots' "brains" are the most and least fit on the basis of the robots' behavior and the current settings of the sliders.

More information can be found here (http://z.cs.utexas.edu/users/nn/nero/about.php), and on the NEAT Wikipedia Page (http://en.wikipedia.org/wiki/NeuroEvolution_of_Augmented_Topologies)

NERO 2.0 has been released fairly recently and can be found from the game's website here (http://z.cs.utexas.edu/users/nn/nero/download.php). I think it would be quite interesting to see what GamingSteve forum members can come up with, and maybe have some battles amongst ourselves!

Nice find Daxx, this seems really interesting and I am not being sarcastic. I think every one should really take this into account, a game with evolving your own team of robots. Neat!!! ;)

Oh yeah, First to post W00t!! ;D
Title: Re: NERO
Post by: Mr. Wizard on January 03, 2008, 03:27:37 pm
Ah, I played it back when it was a in the student showcase for 2006, but I didnt know they had released a sequal. :D
Title: Re: NERO
Post by: Gianavel on January 03, 2008, 07:52:31 pm
Hmm.

Anybody remember the show Odyssey Five?
Title: Re: NERO
Post by: DaMuncha on January 03, 2008, 08:41:11 pm
I installed it, ran it, and got blocks and black squares all over my screen, the window was unresponsive and it crashed.

*throws game in rubbish bin*
Title: Re: NERO
Post by: Daxx on January 04, 2008, 04:08:26 am
I installed it, ran it, and got blocks and black squares all over my screen, the window was unresponsive and it crashed.

Sounds like your computer can't handle it properly.

I've been evolving a team which tries to hang back at the maximum range of the gun and snipes at the enemy. It's quite tricky to get them to do that without them either running away en masse or facehugging. Getting them to handle stuff in the way like walls is going to be even more tricky, I'm warranting.

How's anyone else finding it?
Title: Re: NERO
Post by: Cobra on January 04, 2008, 04:37:31 am
Where can I find system requirements?
Title: Re: NERO
Post by: Slartibartfast on January 04, 2008, 05:03:02 am
I will attempt to install this when I get home.
Title: Re: NERO
Post by: Daxx on January 04, 2008, 05:44:56 am
Where can I find system requirements?

They're pretty low, I'll see if I can hunt them down for you.

EDIT: Actually, I can't find them anywhere. But if my crappy laptop can run it, I'm guessing anyone can.
Title: Re: NERO
Post by: Cocytus on January 04, 2008, 07:07:20 am
If you want I can ask Prof. Stanley (the creator of NEAT) on Monday when I get to class. If he doesn't know I'm sure he can ask his buddies in Austin, TX.
Title: Re: NERO
Post by: Slartibartfast on January 04, 2008, 08:04:05 am
Got it working.  Now I just need to figure it out.
Title: Re: NERO
Post by: Gunner on January 04, 2008, 08:14:49 am
Downloading...
Title: Re: NERO
Post by: Daxx on January 04, 2008, 08:49:07 am
If you want I can ask Prof. Stanley (the creator of NEAT) on Monday when I get to class. If he doesn't know I'm sure he can ask his buddies in Austin, TX.

You know him? That's pretty awesome. :D

This might be an accident, but it's a happy one. My troops have started to take up a cardioid pattern around the enemy, slightly closer at the back. I surmise that if I set up a turret and penalise getting hit I might be able to get them to approach from the rear. Now that's pretty cool, I wasn't able to do that in previous versions.
Title: Re: NERO
Post by: Doctor Z on January 04, 2008, 11:24:55 am
I'm gonna try making guerrilla warfare-using ninjas. Any tips?
Title: Re: NERO
Post by: Neoadept on January 04, 2008, 12:03:25 pm
2.0 came out?  Awesome, now I have a reason to download it again.  See you folks on the battle field.
Title: Re: NERO
Post by: Plasmodia on January 04, 2008, 01:06:15 pm
This looks interesting.. clicks download*
Title: Re: NERO
Post by: Shadowgandor on January 13, 2008, 10:19:02 am
cool :D
Title: Re: NERO
Post by: Daxx on January 14, 2008, 02:35:49 am
Has anyone else gotten any further with this?
Title: Re: NERO
Post by: Poised on January 14, 2008, 01:05:42 pm
Very interesting simulator, but I do have 2 questions, the white slider that determines the distances my NERO's stay from enemies, what do the measuring units represent ? meters ? feet ?
And how far are the NERO's able to shoot from, measured in those same units ?
Title: Re: NERO
Post by: GrapeFruit on January 18, 2008, 04:32:55 am
Pretty awesome stuff...unless it crashes after running smoothly for about 30 minutes
pity


Very interesting simulator, but I do have 2 questions, the white slider that determines the distances my NERO's stay from enemies, what do the measuring units represent ? meters ? feet ?
And how far are the NERO's able to shoot from, measured in those same units ?
That's exactly what I was wondering about...