Human-like Mech AI Decision-making using behaviour trees, followed by a tournament.
Designed solo over a 3 week timeframe and finished on 16/06/2023.
Each student was given code that enables the bots to function, such as shooting,
moving and patrolling. A basic AI was given to us, and our aim was to change it
so it's more human-like. My interpretation of this was an AI that didn't always
land every shot, and wasn't always sure where the others were around the map. However,
it was smart enough to know when to retreat, and when to engage an enemy, based on
heuristic functions that tracked its resources. The tournament was live-streamed,
though since my bot didn't make it to the semis and finals, it wasn't shown.
Researcher
I conducted research on player behaviour in
first-person-shooters, taking notes on the fundamentals.
Programmer
Implementation of the behaviour tree to allow AI
to make decisions based on its environment and resources.
Unity/C#
The primary software used for implentation, and the two prior years
of practice helped me understand how the original AI functioned, so I can tweak it.
Panda BT
Software used to save time writing code in C#. It was fun and easy to
write code in this language. It was interesting to see each decision function change colours based on
the given conditions, since it was a visual way of seeing the brain of the bot.