The goal of this project is to evaluate and improve Neuroevolution as a machine learning technique in a wide variety of complex environments provided by real-time computer games. To date, we have used the TIELT framework to embed the NEAT algorithm in the Unreal Tournament computer game, successfully evolving simple navigation behaviors. Our work has also allowed us to propose a set of improvements for TIELT (Testbed for Integration and Evaluation of Learning Techniques), which is a platform that seeks to provide researchers with a uniform interface to a library of computer game environments.
- Example of initial (random) agent behavior (AVI, MOV)
- Example of evolved agent behavior: the agent moves towards a target on the left wall of the level (AVI, MOV)