Darwinbots Forum

Code center => Darwinbots3 => Topic started by: Numsgil on November 20, 2007, 08:57:53 AM

Title: Physics engine
Post by: Numsgil on November 20, 2007, 08:57:53 AM

I've decided on a physics engine to use.  I was on the fence as to wether I'd use a prebuilt engine or a custom one, but I've decided to at least start with a prebuilt solution.  If I need to later, I can always create a custom physics engine.  Anyway, Chipmunk2D (http://wiki.slembcke.net/main/published/Chipmunk) is what I settled on.  I'm currently working on a .NET wrapper for it.  Bots will be rigid body capsules that can stretch/shrink themselves in limited ways.

You can see a youtube of the stuff Chipmunk can do on youtube (http://www.youtube.com/watch?v=mzdJ2TX8yuk).

Title: Physics engine
Post by: Botsareus on December 20, 2007, 02:15:51 PM

I hope you will not change the physics engine too much so it is easyer to tranlsate robots from earlyer virsions...

Title: Physics engine
Post by: Numsgil on December 20, 2007, 02:19:25 PM

That's not really one of my concerns.

Title: Physics engine
Post by: Botsareus on December 20, 2007, 02:27:59 PM

 that means all the cool old bot's will get lost... Can you please make it a little bit your consern??? :sad:

Title: Physics engine
Post by: MacadamiaNuts on December 20, 2007, 03:35:18 PM

That Chipmunk looks utterly amazing. I've browsed the other YT videos.

Title: Physics engine
Post by: Peter on December 20, 2007, 03:37:00 PM

Well, I don't think it would be very easy to let the old bots work and make something new.

I don't care that much if the older bots don't work very good, or not even work. As long as the tie-phycics work fine in the new version, in the VB-version they are somewhat springy and unpredictable.

If the new version comes out it means we gotta make many, many new bots.

Title: Physics engine
Post by: EricL on December 20, 2007, 03:41:36 PM

Quote from: Peter

As long as the tie-phycics work fine in the new version, in the VB-version they are somewhat springy and unpredictable.

I'm always open to suggestions.  Changing the default inherent springyness of ties is relativly easy.  I would love it if someone would write some test bots using .stifftie and give me feedback on whether we should change the springness over the range of values including the default...

Title: Physics engine
Post by: Numsgil on December 20, 2007, 04:16:18 PM

I think the issues with ties at present is that they're springy at all.  It's like trying to build a building using silly putty.  Or more accurately it's like trying to build a structure using springs

To fix it you'd need to move to a rigid body physics system.  Which is a can of worms I've been opening for most of two years