Gameplay Test Automation with Reinforcement Learning

To ensure high performance for AMD’s graphics cards, the company performs extensive testing on computer game titles. Most gameplay testing is done manually, which results in significant effort and cost expenditure. Thus, this project’s objective is to develop a program capable of learning to automatically play a modern video game. Rather than aiming to optimize performance in the game, the goal is to automate basic actions to explore various graphical scenes the game offers, for the purpose of testing graphics cards. This program will run using reinforcement learning, a type of machine learning focused on enabling a program to learn how to navigate environments by offering it rewards for taking optimal actions. As such, the project’s success would also result in a deeper understanding of reinforcement learning algorithms in practice.

Faculty Supervisor:

Amir-massoud Farahmand

Student:

Partner:

AMD Canada

Discipline:

Computer science

Sector:

Manufacturing; Professional, scientific and technical services

University:

University of Toronto

Program: