AI Behaviour Trees

This project was created as part of a tournament. 48 participants were challenged to program an AI for a mech in an arena shooter style game project. A base FSM AI implementation was provided to participants and they were tasked with designing and implementing a behaviour tree implementation and given the freedom to customize the behaviour in each state: roam, pursue, attack, and flee. The tournament ended with my bot achieving the 1st place position.

My implementation aimed to be more "human-like". I achieved this by customising each of the states to mimic some observed player behaviour in addition to adding some extra behaviours.
More specifically, the bot would:
- Strafe in combat to avoid enemy fire.
- Attempt to flee out of line of sight and restock as quickly as possible when threatened.
- Avoid roaming in high risk areas such as the centre and spawn points.
- Prioritize roaming towards restock points if low on health or ammo.

to add to these, the bot also had a stress value which is based on its current health and ammo count. this value causes the bot to make "bad" decisions in high pressure situations as it affects the bot's decision making abilities.