damn, i will upload it later but at least you will have Screenshots from my interesting project (no one has done it lol). So you will have a clue what's going on there
-
In this image, i have a farm of many ml-agents learning and feeding the NN model their experience.
-
In Hunter's perspective, he's rewarded for catching prey, and prey is rewarded for avoiding hunter, and taking food.
-
i ran this learning for about an hour.
-
here's cumulative reward. You can see over time that prey is getting more and more rewards, it's because he has learned that he should avoid hunter by any means. Hunter's reward stayed the same, because it's very easy to catch a prey and avoid walls.
