BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Tokyo
X-LIC-LOCATION:Asia/Tokyo
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:JST
DTSTART:18871231T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250110T023312Z
LOCATION:Hall B7 (1)\, B Block\, Level 7
DTSTART;TZID=Asia/Tokyo:20241204T145600
DTEND;TZID=Asia/Tokyo:20241204T150800
UID:siggraphasia_SIGGRAPH Asia 2024_sess120_papers_739@linklings.com
SUMMARY:PC-Planner: Physics-Constrained Self-Supervised Learning for Robus
 t Neural Motion Planning with Shape-Aware Distance Function
DESCRIPTION:Technical Papers\n\nXujie Shen, Haocheng Peng, and Zesong Yang
  (State Key Laboratory of CAD&CG, Zhejiang University); Juzhan Xu (Shenzhe
 n University (SZU)); Hujun Bao (State Key Laboratory of CAD&CG, Zhejiang U
 niversity); Ruizhen Hu (Shenzhen University (SZU)); and Zhaopeng Cui (Stat
 e Key Laboratory of CAD&CG, Zhejiang University)\n\nMotion Planning (MP) i
 s a critical challenge in robotics, especially pertinent with the burgeoni
 ng interest in embodied artificial intelligence. Traditional MP methods of
 ten struggle with high-dimensional complexities. Recently neural motion pl
 anners, particularly physics-informed neural planners based on the Eikonal
  equation, have been proposed to overcome the curse of dimensionality. How
 ever, these methods perform poorly in complex scenarios with shaped robots
  due to multiple solutions inherent in the Eikonal equation. %local minima
  of their solutions. To address these issues, this paper presents \pcplann
 er, a novel physics-constrained self-supervised learning framework for rob
 ot motion planning with various shapes in complex environments. To this en
 d, we propose several physical constraints, including monotonic and optima
 l constraints, to stabilize the training process of the neural network wit
 h the Eikonal equation. Additionally, we introduce a novel shape-aware dis
 tance field that considers the robot's shape for efficient collision check
 ing and Ground Truth (GT) speed computation. This field reduces the comput
 ational intensity, and facilitates adaptive motion planning at test time. 
 Experiments in diverse scenarios with different robots demonstrate the sup
 eriority of the proposed method in efficiency and robustness for robot mot
 ion planning, particularly in complex environments.\n\nRegistration Catego
 ry: Full Access, Full Access Supporter\n\nLanguage Format: English Languag
 e\n\nSession Chair: Hao (Richard) Zhang (Simon Fraser University, Amazon)
URL:https://asia.siggraph.org/2024/program/?id=papers_739&sess=sess120
END:VEVENT
END:VCALENDAR