BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Hong_Kong
X-LIC-LOCATION:Asia/Hong_Kong
BEGIN:STANDARD
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
TZNAME:HKT
DTSTART:19911015T033000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20251218T030657Z
LOCATION:Meeting Room S221\, Level 2
DTSTART;TZID=Asia/Hong_Kong:20251215T165100
DTEND;TZID=Asia/Hong_Kong:20251215T170200
UID:siggraphasia_SIGGRAPH Asia 2025_sess115_papers_2447@linklings.com
SUMMARY:FreeMusco: Motion-Free Learning of Latent Control for Morphology-A
 daptive Locomotion in Musculoskeletal Characters
DESCRIPTION:Minkwan Kim and Yoonsang Lee (Hanyang University)\n\nWe propos
 e FreeMusco, a motion-free framework that jointly learns latent representa
 tions and control policies for musculoskeletal characters. By leveraging t
 he musculoskeletal model as a strong prior, our method enables energy-awar
 e and morphology-adaptive locomotion to emerge without motion data. The fr
 amework generalizes across human, non-human, and synthetic morphologies, w
 here distinct energy-efficient strategies naturally appear—for example, qu
 adrupedal gaits in Chimanoid versus bipedal gaits in Humanoid. The latent 
 space and corresponding control policy are constructed from scratch, witho
 ut demonstration, and enable downstream tasks such as goal navigation and 
 path following—representing, to our knowledge, the first motion-free metho
 d to provide such capabilities. FreeMusco learns diverse and physically pl
 ausible locomotion behaviors through model-based reinforcement learning, g
 uided by the locomotion objective that combines control, balancing, and bi
 omechanical terms. To better capture the periodic structure of natural gai
 t, we introduce the temporally averaged loss formulation, which compares s
 imulated and target states over a time window rather than on a per-frame b
 asis. We further encourage behavioral diversity by randomizing target pose
 s and energy levels during training, enabling locomotion to be flexibly mo
 dulated in both form and intensity at runtime. Together, these results dem
 onstrate that versatile and adaptive locomotion control can emerge without
  motion capture, offering a new direction for simulating movement in chara
 cters where data collection is impractical or impossible.\n\nRegistration 
 Category: Full Access, Full Access Supporter\n\nSession Chair: Bo Ren (TMC
 C, College of Computer Science, Nankai University)\n\n
END:VEVENT
END:VCALENDAR