Darwinbots Forum

Bots and Simulations => Bot Tavern => Topic started by: EricL on April 20, 2008, 02:19:21 PM

Title: Seasnake - My Amazing Multibot
Post by: EricL on April 20, 2008, 02:19:21 PM
I'm working on a really cool multibot and debugging the multibot features as I go.  It's become clear to me that multibots have tremendous inherent advantages that would totally crush the F1 field and own IM if we just got the bugs out of ties and such so that is what I am doing, building this bot as I go a test case.

I can't post a sim yet since it requires some fixes in 2.43.1j, but here's a screen shot of couple of them near some veggies.
Title: Seasnake - My Amazing Multibot
Post by: The_Duck on April 20, 2008, 02:40:29 PM
Looks nice; what fixes did you make?
Title: Seasnake - My Amazing Multibot
Post by: EricL on April 20, 2008, 03:28:02 PM
Quote from: The_Duck
Looks nice; what fixes did you make?

I'll document them when I drop the release, but...

There were problems with both .fixang and .fixlen.  There was duplicate code in two different areas each trying to apply the values differently, resulting in unpredictable behaviour such as tie angle occsilations.

Even the stiffest tie was still very springy, so I increasd the range of values that matter on .stifftie to 1-100 from 1-40 (essentially adding the ability to specify higher degrees of tie stiffness than before - values 1-40 equate to the same stiffness as before, values 41-100 equate to greater stiffness) and now MOD positive values to spread mutation probabiltiy equally across the range instead of pegging anythign above the max and the max stiffness (as values > 40 did before).  All negative values equate to the minimal tie stiffness of 1.   A 0 value does not change anything as before - hardened ties remain at whatever stiffness they are at.

I made the changes to .fixang discussed here (http://www.darwinbots.com/Forum/index.php?showtopic=2663) to allow bots the option to specify that the tie angle should rotate freely by storing a negative number.

I also fixed some bugs with trefvars not going away in certain cases when ties break as well as erasing the trefvars when .readtie is non-zero but invalid.  This lets bot check if ties for which they know the tie port still exist or not.
 
I suspect there are furhter issues with tie sharing and bot order of operation.  I'm investigating those now.
Title: Seasnake - My Amazing Multibot
Post by: Moonfisher on April 20, 2008, 05:35:50 PM
Sweet, tried to make a multibot earlyer but gave up because there was too much stuff going on with the tie angles and such...
Title: Seasnake - My Amazing Multibot
Post by: EricL on April 22, 2008, 11:42:42 PM
Here is a sim which demonstrates the V1 version of this multibot.

Requires version 2.43.1j or above.