The founding purpose of Artificial Intelligence (AI) is to build human-level (or stronger) intelligence in an artificial entity. Video games provide a suitable testbed as they are rich enough to be challenging for AI algorithms yet simple enough to get traction and make progress. From the application perspective, procedural content and behavior generation is becoming critical in making the development of AAA titles tractable. Video game companies are interested in generating high-fidelity non-playable-character (NPC) behavior including dialogue, emotional reactions and routine activities. In this course we will survey classic and state-of-the-art AI methods used in video games. We will consider the theoretical foundations (e.g., Markov decision processes) as well as software technology involved. Certain lectures may be given by representatives of the industry (e.g., BioWare staff). Students in the class will pick a specific aspect of NPC behavior and work in teams on implementing it. Some projects may be extended into a thesis after the course.
Course syllabus (PDF, 124kb)
This course pursues the following objectives: (i) gaining an understanding of current challenges in video game AI; (ii) learning classic and state-of-the-art AI algorithms applicable to the challenges; (iii) presenting and critiquing those algorithms in class; (iv) gaining hands-on experience and working knowledge of some of the algorithms; (v) doing original research n the form of a team-based term project; (vi) writing a conference-level report on the research; (vii) gaining ideas for your M.Sc./Ph.D. degree project.
Most lectures will be of the seminar type where we will discuss papers and case studies. Everyone in class will read the papers or play the game ahead of time. One student will present the paper/game in class with slides and/or live demonstrations. Everybody else will hand-in their written review of the paper at the beginning of the class. Following each presentation, we will have a discussion led by the presenter. The presentations and the written reviews will be marked by the instructor. Marking templates and review/presentation guides will be provided.
The students will form 2-3 person teams and work together throughout the term. They will propose a research topic to explore and, upon instructor's approval, follow several milestones (e.g., a midterm report, an in-class presentation, a final report, etc.) until the end of the term. Guidelines for all of these will be provided in class.