GAIG Game AI Research Group @ QMUL

Alexander Dockhorn

I studied Computer Science at the Otto-von-Guericke University (OvGU) in Magdeburg (Germany) and University of Abertay in Dundee (Scotland). Back in Magdeburg I finished my Master in Computer Science and after which I joined the Computational Intelligence research group lead by Prof. Rudolf Kruse and Prof. Sanaz Mostaghim. After finishing my PhD thesis on the topics forward model learning and predictive state determinization I joined the Game AI Research Group as a Postdoctoral Research Assistant. My current research interests are intelligent data analysis and more generalised methods for forward model learning.

Abstract Forward Models for Modern Games

This project aims to provide the games industry with access to the latest and most proficient Game AI methods. Statistical Forward Planning (SFP) techniques, such as Monte Carlo Tree Search (MCTS) or Rolling Horizon Evolutionary Algorithms (RHEA), have recently achieved remarkable performance in games research. This project addresses the main reasons behind the small uptake of SFP methods in the games industry: the lack of fast and reliable Forward Models (FM) that can be abstracted for its use by SFP algorithms in modern video-games. SFPneedsanFMtowork,butcomplexmodelsareexpensiveto use without introducing abstractions or simplifications, which also make the model inaccurate.

Hearthstone AI - International Research Competition

I am the organiser of the international research competition on Hearthstone AI, which was part of the IEEE Conference on Computational Intelligence and Games 2018-2020. The competition focuses on the development of autonomous Hearthstone agents and tested the agent’s skill on multiple game playing tasks.

Competition Webpage

GAIG Publications