Info
This post was imported from a personal note. It may contain inside jokes, streams of consciousness, errors, and other nonsense.
Okay, I implemented a BraitenBoid. So far not working. Like, at all.
I tried changing from having the sensors detect all nearby food sources to just the closest. That improved the speed from 5-8 seconds for 10k steps down to 4-6 seconds at least.
The regular Boids will usually get better results and stay there after 5-6 generations. This one just seems to luck out and get fewer than 10k steps here and there but doesn’t actually improve at all.
Boo.
What do the weights mean in this case? I’m still using the same turn output, instead of left and right motors. One thing at a time.
What would I expect to perform well for the weights?
w0 (bias to speed) - Positive to tend to keep moving.
w1 (bias to turn) - Ideally zero but it usually doesn't pan out that way.
w2 (left food detect to speed) - Slowing down might actually be good. Negative?
w3 (left food detect to turn) - Negative so when it's activated it'll turn left.
w4 (right food detect to speed) - Maybe slow down? So a little negative?
w5 (right food detect to turn) - Positive so when it's activated it'll turn right.
What am I getting? Or at least, what did I get this time? I’m not confident it’s consistent.
Generation: 30
Steps: 11334
Fitness scores, weights:
4 0.02 -0.49 -0.55 0.27 0.90 0.28
2 0.12 -0.48 -0.50 0.20 0.90 0.11
2 0.35 -0.54 -0.42 0.13 0.93 0.20
1 0.03 -0.41 -0.45 0.09 0.97 0.29
1 0.29 -0.85 -0.58 0.39 1.00 0.30
1 0.36 -0.70 -0.56 0.10 0.94 0.17
1 0.10 -0.51 -0.52 0.12 0.90 0.31
1 0.16 -0.44 -0.48 0.21 0.84 0.29
0 -0.04 -0.49 -0.37 0.11 0.99 0.25
0 0.31 -0.79 -0.52 0.36 0.90 0.30
Another attempt at 30 generations (well, 31 after running visualize):
Generation: 30
Steps: 10260
Fitness scores, weights:
5 0.09 0.60 0.53 0.71 -0.21 -0.19
2 0.16 0.88 0.31 0.58 -0.18 -0.01
2 0.10 0.78 0.29 0.64 -0.26 0.11
1 0.21 0.79 0.13 0.65 -0.26 -0.04
1 0.21 0.68 0.61 0.78 -0.26 -0.16
0 -0.02 0.98 0.22 0.59 -0.31 0.18
0 0.09 0.65 0.47 0.75 -0.33 0.03
0 -0.05 0.91 0.30 0.60 -0.35 0.06
0 -0.05 0.61 0.38 0.67 -0.46 -0.09
0 0.13 0.94 0.21 0.53 -0.38 0.07
They mostly just all going in circles.
One more just in case:
Generation: 30
Steps: 14705
Fitness scores, weights:
1 1.00 0.35 -0.58 0.05 0.37 -0.02
1 0.97 0.49 -0.38 -0.10 0.48 0.03
1 0.95 0.15 -0.44 -0.21 0.42 0.02
1 0.89 0.42 -0.40 -0.03 0.45 -0.06
1 0.90 0.02 -0.36 0.08 0.55 0.02
1 0.94 0.48 -0.47 -0.04 0.49 0.10
0 0.96 0.03 -0.23 0.11 0.58 -0.08
0 1.00 0.42 -0.48 0.04 0.45 -0.04
0 0.97 0.12 -0.58 -0.27 0.46 0.08
0 0.97 0.53 -0.59 0.09 0.39 0.08
Maybe this one seems a little better? One of them even looked like it steered toward a bit of food.
Well, this is star far from what I was hoping for. I’ll need to experiment more with this tomorrow. My code for the sensor positions is really shaky. I’ll test it tomorrow.
Maybe literally write some tests for it, actually.