This is something i need to code soon, however i was inspired by some other post (this is probably next step to that other question).
So as example i use this situation:
- i have event that happens at 1700 its “5 o clock, time for tea”
- i have 100 AI bots in small town. If they all go exactly at 1700 it will look scripted and unnatural.
- so i want to make normal distribution. And they all know 5 o clock always happens at 1700 in advance.
Now i want them to find slot when they go for that tea, but result needs to follow normal distribution.
My first idea was to “brute force” this:
I split that 1630 to 1730 into 100 intervals, and each bots rolls number and checks if its belo normal distribution level, then goes for tea or not. But since they leave pool of active (thirsty) bots, resulting curve is deformed.
Second idea is to roll in advance (like every bot rolls when it goes for tea (this should be even distribution). Then i check it against normal distribution, and make extras roll again. However this is ugly solution.
Is there any nice solution for it. And bonus points if i could just draw distribution function and use that.
ps.
Yes i am advocate for Deep Ai to research stuff lik this, it usually gives good leads, and not so good solutions. However this time it went all “Derpy Derp Aii” route, just mutations of my first idea, then circular fixing some aspects and ruining other.
So far i found this:
" Box–Muller transform" - creates normal distribution, and its elegant. However still making custom distribution from function is unresolved.
