Thu, January 2, 10:00 AM
45 MINUTES
Animation Synthesis using Machine Learning

Automating character control in vir-tual environments is an increasing-ly popular approach for synthesiz-ing procedural animation in video games. This requires a method that outputs, for each timestep, simulation actuation parameters such as joint torques or angles such that the character performs some desired move-ment. This poses a continuous optimization problem with high dimensionality and a large number of physics- based constraints.This talk gives an overview of AI research for synthesizing procedural animations in games. The main focus of the talk will be on physics-based control; however, some of the kinematic-based methods will also be covered. The topic can be useful for people interested in online/offline optimi- zation, supervised learning and reinforcement learning.

Amin Babadi

Ph.D. candidate at Department of Computer Science, Aalto University

Amin is a Ph.D. candidate at Department of Computer Science, Aalto University, Finland. He works under supervision of Prof. Perttu Hämäläinen. Amin is also a visiting researcher at Imager lab, University of British Columbia, Canada, where he works with Prof. Michiel van de Panne. His current research focuses on developing efficient, creative movement artificial intelligence (AI) for physically-simulated characters in multi-agent settings.Prior to his Ph.D., Amin had ten years of experience in the video game industry. Specifically, he has worked on several commercial games from various genres including first-person shooter, two-player football, and classic adventure. In these projects, he has been responsible for different programming disciplines including AI, animation, gameplay, and physics.