BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Australia/Melbourne
X-LIC-LOCATION:Australia/Melbourne
BEGIN:DAYLIGHT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
TZNAME:AEDT
DTSTART:19721003T020000
RRULE:FREQ=YEARLY;BYMONTH=4;BYDAY=1SU
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:19721003T020000
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
TZNAME:AEST
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240214T070242Z
LOCATION:Meeting Room C4.11\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231212T142500
DTEND;TZID=Australia/Melbourne:20231212T144000
UID:siggraphasia_SIGGRAPH Asia 2023_sess120_papers_841@linklings.com
SUMMARY:Neural Categorical Priors for Physics-Based Character Control
DESCRIPTION:Technical Papers\n\nQingxu Zhu, He Zhang, Mengting Lan, and Le
 i Han (Tencent)\n\nRecent advances in learning reusable motion priors have
  demonstrated their effectiveness in generating naturalistic behaviors. In
  this paper, we propose a new learning framework in this paradigm for cont
 rolling physics-based characters with significantly improved motion qualit
 y and diversity over existing state-of-the-art methods. The proposed metho
 d uses reinforcement learning (RL) to initially track and imitate life-lik
 e movements from unstructured motion clips using the discrete information 
 bottleneck, as adopted in the Vector Quantized Variational AutoEncoder (VQ
 -VAE). This structure compresses the most relevant information from the mo
 tion clips into a compact yet informative latent space, i.e., a discrete s
 pace over vector quantized codes. By sampling codes in the space from a tr
 ained categorical prior distribution, high-quality life-like behaviors can
  be generated, similar to the usage of VQ-VAE in computer vision. Although
  this prior distribution can be trained with the supervision of the encode
 r's output, it follows the original motion clip distribution in the datase
 t and could lead to imbalanced behaviors in our setting. To address the is
 sue, we further propose a technique named prior shifting to adjust the pri
 or distribution using curiosity-driven RL. The outcome distribution is dem
 onstrated to offer sufficient behavioral diversity and significantly facil
 itates upper-level policy learning for downstream tasks. We conduct compre
 hensive experiments using humanoid characters on two challenging downstrea
 m tasks, sword-shield striking and two-player boxing game. Our results dem
 onstrate that the proposed framework is capable of controlling the charact
 er to perform considerably high-quality movements in terms of behavioral s
 trategies, diversity, and realism.\n\nRegistration Category: Full Access\n
 \nSession Chair: Jungdam Won (Seoul National University)
URL:https://asia.siggraph.org/2023/full-program?id=papers_841&sess=sess120
END:VEVENT
END:VCALENDAR