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:20260114T163655Z
LOCATION:Meeting Room C4.11\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231212T140000
DTEND;TZID=Australia/Melbourne:20231212T141000
UID:siggraphasia_SIGGRAPH Asia 2023_sess120_papers_619@linklings.com
SUMMARY:C·ASE: Learning Conditional Adversarial Skill Embeddings for Physi
 cs-based Characters
DESCRIPTION:Zhiyang Dou (The University of Hong Kong (HKU)), Xuelin Chen a
 nd Qingnan Fan (Tencent AI Lab), Taku Komura (University of Hong Kong), an
 d Wenping Wang (Texas A&M University)\n\nWe present C·ASE, an efficient an
 d effective framework that learns conditional Adversarial Skill Embeddings
  for physics-based characters. Our physically simulated character can lear
 n a diverse repertoire of skills while providing controllability in the fo
 rm of direct manipulation of the skills to be performed. C·ASE divides the
  heterogeneous skill motions into distinct subsets containing homogeneous 
 samples for training a low-level conditional model to learn conditional be
 havior distribution. The skill-conditioned imitation learning naturally of
 fers explicit control over the character's skills after training. The trai
 ning course incorporates the focal skill sampling, skeletal residual force
 s, and element-wise feature masking to balance diverse skills of varying c
 omplexities, mitigate dynamics mismatch to master agile motions and captur
 e more general behavior characteristics, respectively. Once trained, the c
 onditional model can produce highly diverse and realistic skills, outperfo
 rming state-of-the-art models, and can be repurposed in various downstream
  tasks. In particular, the explicit skill control handle allows a high-lev
 el policy or user to direct the character with desired skill specification
 s, which we demonstrate is advantageous for interactive character animatio
 n.\n\nRegistration Category: Full Access\n\nSession Chair: Jungdam Won (Se
 oul National University)\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_619&sess=sess120
END:VEVENT
END:VCALENDAR