Data Science Rzeszów #8: AI in video games by Rafał Tyl
All are welcome to attend Data Science Rzeszów Meetup, on the agenda:
AI in video games
Most video games feature some kind of non-player characters that the player can interact with – allies, enemies or just ambient crowd. No matter the type of the NPC in question, they all require proper algorithms to govern their behavior in the most believable and fun way possible – and all of that in real time, with a computation budget that’s often severely restricted. Another difficulty arises from the nature of game production process – various assumptions and design choices change incessantly and AI systems have to be flexible enough to adapt to these changes. These requirements pose a serious challenge for machine learning methods, causing most developers to ignore those methods altogether.
This presentation describes key components of a complete game AI system along with a brief description of algorithms and methods typically used in each of those. It also identifies various challenges related to using machine learning methods in such systems, while still pointing to areas that could easily benefit from those techniques
Rafał Tyl is a game programmer, primarily focused on artificial intelligence. Graduated from the Faculty of Mechatronics of Warsaw University of Technology, specializing in image processing, but switched fields to game programming soon afterwards. He is experienced in traffic simulation for mobile games and AI for real time strategies (e. g. Ancestors Legacy by Destructive Creations). Employed by QED Software since 2017, where he co-created Grail framework. Currently working on a tactical game titled Tactical Troops: Anthracite Shift and Sensei – an ML-based game analytics system.
Written by
