3D Image Representation for Gameplay Automation

AMD graphics cards are extensively used for computer games, and testing their performance involves manually playing various titles on multiple configurations. Automated gameplay can save time and effort during this process. However, using game images as input presents challenges for reinforcement learning (RL) agents, specifically in player navigation and movement. This internship aims to develop solutions to convert 3D images or videos from modern AAA game titles into more accessible representations for ML or RL models, focusing on navigation automation. The objective is to optimize game/image features for navigation and exploration, making the solution adaptable to different games, with a primary focus on first-person, open-world titles. The solution should operate in real time, without interfering with gameplay actions or other simultaneous processes.

Faculty Supervisor:

Steve Engels

Student:

Partner:

AMD Canada

Discipline:

Computer science

Sector:

Manufacturing; Professional, scientific and technical services

University:

University of Toronto

Program: