BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Seoul
X-LIC-LOCATION:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:18871231T000000
DTSTART:19881009T020000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230103T035307Z
LOCATION:Room 324\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221206T140000
DTEND;TZID=Asia/Seoul:20221206T153000
UID:siggraphasia_SIGGRAPH Asia 2022_sess154_papers_144@linklings.com
SUMMARY:Control VAE: Model-Based Learning of Generative Controllers for Ph
 ysics-Based Characters
DESCRIPTION:Technical Communications, Technical Papers\n\nControl VAE: Mod
 el-Based Learning of Generative Controllers for Physics-Based Characters\n
 \nYao, Song, Chen, Liu\n\nIn this paper, we introduce Control VAE, a novel
  model-based framework for learning generative motion control policies bas
 ed on variational autoencoders (VAE). Our framework can learn a rich and f
 lexible latent representation of skills and a skill-conditioned generative
  control policy from a diverse set of unorganized motion sequences, which 
 enables the generation of realistic human behaviors by sampling in the lat
 ent space and allows high-level control policies to reuse the learned skil
 ls to accomplish a variety of downstream tasks. In the training of Control
  VAE, we employ a learnable world model to realize direct supervision of t
 he latent space and the control policy. This world model effectively captu
 res the unknown dynamics of the simulation system, enabling efficient mode
 l-based learning of high-level downstream tasks. We also learn a state-con
 ditional prior distribution in the VAE-based generative control policy, wh
 ich generates a skill embedding that outperforms the non-conditional prior
 s in downstream tasks. We demonstrate the effectiveness of Control VAE usi
 ng a diverse set of tasks, which allows realistic and interactive control 
 of the simulated characters.\n\nRegistration Category: FULL ACCESS, ON-DEM
 AND ACCESS\n\nLanguage: ENGLISH\n\nFormat: IN-PERSON, ON-DEMAND
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_144&sess=sess15
 4
END:VEVENT
END:VCALENDAR