Obstacle Learning AI

OK. I’m just learning some stuff and testing some stuff right now. I’m trying to make and AI that goes through an obstacle course to reach it’s goal. I have the basics down, but I need this AI to create and save solutions in a sort of memory, so that it may use them further on. Can someone help me with this?

3 Likes

Hey, your AI/CPU specialist here, I’m having trouble understanding what you need help with.

1 Like

Hi. This AI will start off slow in the obstacle course, just moving one way at a time. When it gets past an obstacle course, it will save its movements in a database, so when it encounters this again it’ll know what to. Eventually, the AI gets so proficient that they speed through the obstacles like Lightning Mc Queen. That’s the gist of it.

2 Likes

Two problems with different solutions:

  1. You need help making a “smart AI”? What I would do for this is use a system of raycasts. When the raycasts hit said dodging obstacle, your object reacts.

Example:

An object is running along a track when it approaches a block in the way. A raycast will detect this block, and when it does, it will trigger the object to jump.

  1. You want to be able to “unlock” character movements? What I would do for this, is only have the raycast actually cast the ray once a save behavior unlocks the feature.

Example:

Your object is moving in a level too high for it, so it falls into spikes it dies. When it goes back to the correct level, it runs through an invisible block that saves a “1” value into a select save function called ‘UnlockSpike’. The object itself will ‘read’ the value of the save and if it is equal to “1”, then the raycast starts casting the ray.

2 Likes

I think what he wants is to creat a nerual network

He is saying like the Hello Neighbor code when you do a plan to get to basement in hello neighbor and he catches you he remembers what you did (like real life) and you gotta do another plan

Like this
Screenshot 2021-03-10 102247
He wants it that so when the AI does something it learns it and can use it later on

1 Like

I dont think this is doable just because of the lack of features that flowlab provide for anything that thinks on its own so i would check out the unity game engine and learn how to use that for this kind of perpouse

1 Like

I remember sketching out something like this in my planning book but never got around to it since there was a flowjam. Let me see if I can find it.

1 Like