next »

[motorherp]

[Edit - 20/11]

Righto folks.  I'm having to put in some late nights at work at the moment and my last minute holiday is coming up in a few days so realistically I don't think I'm going to have any more time to put into this.  Therefore I'll just post up what I have now.  There's not really much to it, but hey, something is better than nothing right?  Click the link below to play.  Enjoy 

British Bulldogs

---------------------------------------------------------------------------------------------
[Original Post]


I'll take a crack at this.  Elbow kids in the face and run until you vomit, aka British Bulldogs

[capn.lee]

nice

[felipezille]

expecting it already :D

[Panurge]

Nice! Would look cool with an actual playground backdrop and maybe a Grange Hill vibe...

[motorherp]

Hey Panurge, cheers for the feedback.  I did think about setting it on the playground too but sadly I'm really not much of an artist and would struggle to make it look anything better than a flat and boring grey plane.  I found it easier to get results with grass so decided to set it in on the school field instead.

I'm really liking the Grange Hill idea.  One thing I was thinking to give the game more character would be to have the kids throwing playground taunts and insults at each other using little speech bubbles above their heads from time to time.  Grange Hill would be an excellent place to harvest some cool phrases.  I don't know if I'll get that far though.

[motorherp]

Haven't managed to do any work on this all week but started making some more progress today.  I've written a neural-net AI and genetic algorithm to train them up.  I've not done this before so it'll be interesting to see how it turns out.  Just running my trainer now, will probably leave it running over night and see what I end up with in the morning.  Just hope nothing crashes 

[whilefun]

Great idea! Looking forward to seeing what you come up with 

[motorherp]

So the good new is that my neural net trainer has now been running for about 16 hours without crashing.  The bad new is that although my little guys are showing improvements and are clearly at least trying to play the game, they're still not all that clever and their behavior is still fairly erratic.

So far each training session has been set up to run for a maximum of 20 seconds although many of them now finish early when all the survivors are either caught by bulldogs or make it to the safe zone.  Therefore my guys are at least on their 2880th generation but I'm predicting more like 4000 generations (I really should have put a counter in  ).  Being the first time I've done this I've really no idea how long it should take to get good behavior or whether I need to start tweaking the genetic algorithm parameters.

I'm not ready to give up yet though.  I think what I'm going to do next is save off this generation then refactor my code so the neural net trainer can zip through games as quickly as my processor can handle rather than running the games in real time.  That should allow me to progress through each generation much quicker.

I'm thinking its going to pretty unlikely I'll finish this game in time for the deadline now though since I booked a last minute holiday last night for next week so I've only realistically got a few more days to work on this around my day job and preparing for my hols.


Edit - Just had another idea too.  At the moment my neural net inputs are things like distance and direction to nearest bulldogs/survivors and distance from safe zone etc, and my outputs are simply vertical and horizontal move directions and whether to sprint.  I think one thing that will take the system a long time is that I'm expecting it to evolve tactical thinking and basic motion control to achieve its goals at the same time.  What I might do is tinker with having higher level outputs such as 'follow closest target' or 'move towards safe zone' and thus freeing up the neural net to only have to evolve tactical decision making.

[motorherp]

Sweet, its now chugging along at about 10 generations per second or about 200 times faster 

[motorherp]

Righto folks.  I'm having to put in some late nights at work at the moment and my last minute holiday is coming up in a few days so realistically I don't think I'm going to have any more time to put into this.  Therefore I'll just post up what I have now.  There's not really much to it, but hey, something is better than nothing right?  Click the link below to play.  Enjoy 

British Bulldogs

For anyone interested, I had to abandon the neural net idea in the end  .  I'd managed to find and fix a couple of bugs in it over the weekend and after doing a little more research and tweaking my parameters I was starting to make a little progress with evolving some behaviors but it was just going to take me too long to get it to where I wanted.  Therefore I quickly replaced it with some hand-coded AI just to get something out.  It's not the best but it kinda does the job.  The evolving neural net thing was really interesting though, I think I'll have another play with that at some point in the future.

[Fifth]

Aw, sorry to hear you had to abandon the neural net.  That sounded like it was really fun to play with, and I wanted to see how it would have turned out.

[motorherp]

Yeh its a really cool little toy, I'll definitely have another tinker with it in the future.  I'm not experienced at all with neural nets or genetic algorithms though and also I'm pretty convinced there's still some bugs in it too so there was no way I was going to get something workable in time.  What I'm thinking though is that it would be really neat to make a kinda sandbox god sim with it in the future if I work out the kinks.