Grant: Engage Grants Program

As a leading computer game developer, Electronic Arts (EA) Canada wants to deliver acceptable looking animations at interactive speeds (typically 30 or 60 frames per second). Computer games display animations by first updating the game state (e.g., by updating the position of every character in the scene), and then rendering the game state by converting information about the game state into a picture. EA Canada personnel routinely devise procedures (parts of computer game programs) that simplify the updating and rendering processes in order to speed them up. For each part of updating and rendering, the challenge is to make decisions about whether to use a slower method that gives better animation quality or a faster method that gives poorer quality. Making these decisions is time consuming so EA Canada would like to automate them. Since game players are the final judges of animation quality, they would like a calculated measure that mimics human perception of quality. Such a measure could be implemented in software to enable a computer game to efficiently balance animation quality and real-time performance. The main objective of the project is to find a mathematical or algorithmic quality measure that mimics the human perception of the quality of animation. The particular focus will be on measures that perform well in the context of evaluating the same animation clip with varying types and amounts of introduced errors. Some relevant types of errors are jitter/pop, offset errors, interpenetration of geometric models, and loss of motion smoothness (due to jumps in velocity). Ideally, such a quality measure would lend itself to finding thresholds on the amount of each of these types of errors, or combinations of these errors, where animation quality goes from acceptable to unacceptable.