Character animation plays a key role in Games, Virtual Reality (VR), Augmented Reality (AR) applications. Locomotion such as walking and running, enable characters freely navigate in the virtual environment. In this talk, I will present a collection of classic and state-of-the-art approaches for creating locomotion controllers, simulating locomotion in physics environment, and adding styles to the movements. I will start from kinematic methods to create locomotion controllers for biped and quadrupled characters navigating on different terrains, particularly suitable for the game environment. I will further cover locomotion simulations with physics engine, by learning policies from interactions with the environment under the reinforcement learning framework. Locomoting characters can thus adapt motions to external forces and environmental perturbations. Multiple motion style transfer methods will also be discussed to add more diversity and expressiveness to the locomotion in the virtual world.
Yingying Wang is an Assistant Professor at Computing And Software department of McMaster University. Her research interest is mainly in Character Animation, Graphics and Machine Learning. Before joining McMaster University, Yingying was a researcher at Snap Inc., focusing on novel animation research and applications for mobile devices. Previously Yingying worked on gesture synthesis, hand motion capture, motion style transfer, motion retrieval and 3D human pose estimation, which leads to multiple paper and patent publications. Yingying is also an active contributor to several open-source projects.