Anton Debner

Background

About Anton Debner

I have recently finished my PhD in the area of Machine Learning (ML) and Game Development. My research focused on scalability of Deep Reinforcement Learning (DRL) in the context of Game Engines. I believe DRL is the technology that will make game AI more human-like, interesting, fun and creative. For example, DRL can be used to cheaply improve playtesting coverage and to even score a procedurally generated level based on diversity of skills the DRL-agent had to use to solve it. Characters with DRL-trained physics based controls can stumble in difficult terrain, access difficult-to-get-to locations and learn to push other characters around, creating new kinds of game mechanics even for indie games.Outside of DRL, I have been working with simulators/game engines to produce virtual sensor data for autonomous vehicles, industrial warehouses and production plants. The data might include positional data, RGB and RGB-d images and even LiDAR point clouds. This data can then be used, for example, to train or test ML models and finally to transfer the model to real life use.I also have over five years of experience creating and running automatic grading systems for programming courses. These vary from beginner-level programming exercises, to automatically graded free-form UML diagrams and detecting concurrency issues in multi-threaded programs.In my free time, I like to dabble with coding and game-related projects.See my game-related projects and playable demos at:Website: https://debner.fi/GitHub: https://github.com/debnera