The Hunt : Reinforcement learning for studying animal behaviour with Unreal Engine 5
- Category: Games and Reinforcement Learning
- Language: Blueprint and C++
- Date: Still in progress
Introduction
The goal of this project is to compare the result of a simulating between using Behavior Tree and using Machine Learning methods (Neural Network, Reinforcement Learning, genetic AI algorithm...) to describe the behaviour of actors. Which is the best? At what cost? .
To do this, I create an open-world (on Unreal Engine 5) with differents types of animal with unique behaviour :
- Fearful (Deer) : They flee in front of any animal, and try to stay grouped
- Predatory (wolf) : They hunt other animals.
Animals will learn to hunt (predict and react). Each actor will have to manage 3 variables: stamina, hunger, health which will influence the way they behave. Deer needs to eat plants and Wolf needs to eat deer! For now, I start working on RL algorithms with the Mind Maker plugin.
Opportunity for development
This project focuses on research. I will develop it according to my results. Why not add other types of animals and a more complex environment.