Create a system for generating fit objects in a simulated environment. The focus will be on a game environment, and a key takeaway of the project will be how the methods applied may be used in other applications and other simulations.
The environment will be a simple action game called SOL, and the generated content will be characters in the game. The fitness of the generated characters will be based on how interesting/fun they are to play.
The method to accomplish this task will be based on generation and evaluation of characters. Generation by search and evaluation by simulation will be researched. A variant of adaptive stress testing will be tested for this purpose. Other methods that may constitute a solution may include genetic algorithms and other machine learning and reinforcement learning approaches.
Evaluation of simulations will be done by defined metrics or more sophisticated methods.