Thesis Project
The question my thesis projects asked was how best to recreate human behaviour with artificial intelligence in a 2D top down wave shooter game?
To answer this two different agents were made with the goal to play a game with as humanlike behaviour as possible. One agent controlled with a Behaviour tree and one controlled by an LSTM neural network trained using reinforcement learning. When they were ready testers played with both agents one after another, except that the testers were told that one of the agents were an AI and one was another human player. The tester was not told which were which and they had to decide which they thought were the agent. The goal of this was to find out which of the two used techniques were best at imitating human player behaviour. The study was fairly limited in scope, while it would be interesting to test more techniques than just these two, time was a limiting factor and we had to settle with just these two.