Reinforcement Learning for All: An Implementation Using Unreal Engine Blueprint
Reece A. Boyd, Salvador E. Barbosa · 2017
Game engines, like Unreal, Unity, and Cryengine, provide cutting edge graphics, sophisticated physics modeling, and integrated audio, greatly simplifying game development. These advanced features are often accessible through visual scripting interfaces used in rapid prototyping and by non-programmers. The goal of this research was to demonstrate that these tools can support implementation of artificial intelligence techniques, such as reinforcement learning, that have the potential to yield dynamic characters that are not pre-programmed, but rather learn their behavior via algorithms. Its novelties are the implementation of a Q-Learning bot, created in Unreal Engine's visual scripting tool, known as Blueprint.