Menagerie
Fuzzy Brain Behavior Tree AI
The AI in Menagerie was definitely the most advanced I have created. I came up with the idea of combining Fuzzy Brains with Behavior Trees. The result is an interesting, but flawed, product. I also created the music, and the implementation of the music as a dynamically changing thing using FMOD.
This project was created as part of the school curriculum.
Fuzzy Brains
I was frustrated with games where the enemies are stupid to a fault, and from this, the idea came to create a fuzzy brain behavior tree.
The concept revolves around the AI memory. It consists of a series sensory inputs that the AI weighs to make decisions.
In this video from the beta, we can see the effects of this brain in action. The AI goes to patrol, but for a moment it hears something. However, it's not a strong enough signal to warrant deeper investigation, so it lets it go.
To illustrate how the fuzzy brain works, imagine the following happening approximately simultaneously:
- Heard glass shattering nearby
- Glimpsed the player in the distance
- Saw an open door that should be closed
In this case, a simpler AI might deal with one of them and then forget anything ever happened, but with a system like this, everything is accounted for. In order:
- The inputs are weighed based on signal strength, proximity, enemy state, and signal persistence.
- The signals are investigated in order.
So in the case above, it might first run to check what happened with the glass, and then, seeing that there is no one there, it would investigate the fact that it saw a glimpse of the player, and so on.
The methods of investigation change depending on how strong the signal is. For a weaker signal it may be enough to get closer and try to see the source from afar.
In the end, this never reached its full potential because of the time constraints and competing responsibilities.