Survival - Week 0

Survival is aimed to be an AI based experiment using Unity's ml agents plugin and C# to see how efficiently an agent can be trained in a dynamic world.

ML agents is a plugin for unity that supports two types of machine learning methods :- Reinforcement Learning and Imitation Learning.

https://github.com/Unity-Technologies/ml-agents/blob/main/docs/ML-Agents-Overview.md

For this project I will be using reinforcement learning as I want the agent to learn at runtime without any external stimuli.

For more information about reinforcement vs imitation learning :-

https://deeplearninguniversity.com/imitation-learning-and-reinforcement-learning-whats-the-difference/

Reinforcement learning has a lot of different learning algorithms, ML agents uses one such algorithm that is called Proximal Policy Optimization.

More information on PPO can be found https://jonathan-hui.medium.com/rl-proximal-policy-optimization-ppo-explained-77f014ec3f12

The main goals of the project is to give the agent multiple actions to do in a dynamic environment to keep itself alive.