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:20240214T070250Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231215T133000
DTEND;TZID=Australia/Melbourne:20231215T134500
UID:siggraphasia_SIGGRAPH Asia 2023_sess157_papers_419@linklings.com
SUMMARY:Neural Packing: from Visual Sensing to Reinforcement Learning
DESCRIPTION:Technical Papers, TOG\n\nJuzhan Xu (Shenzhen University), Ming
 lun Gong (University of Guelph), Hao Zhang (Simon Fraser University), and 
 Hui Huang and Ruizhen Hu (Shenzhen University)\n\nWe present a novel learn
 ing framework to solve the transport-and-packing (TAP) problem in 3D. It c
 onstitutes a full solution pipeline from partial observations of input obj
 ects via RGBD sensing and recognition to final box placement, via robotic 
 motion planning, to arrive at a compact packing in a target container. The
  technical core of our method is a neural network for TAP, trained via rei
 nforcement learning (RL), to solve the NP-hard combinatorial optimization 
 problem. Our network simultaneously selects an object to pack and determin
 es the final packing location, based on a judicious encoding of the contin
 uously evolving states of partially observed source objects and available 
 spaces in the target container, using separate encoders both enabled with 
 attention mechanisms. The encoded feature vectors are employed to compute 
 the matching scores and feasibility masks of different pairings of box sel
 ection and available space configuration for packing strategy optimization
 . Extensive experiments, including ablation studies and physical packing e
 xecution by a real robot (Universal Robot UR5e), are conducted to evaluate
  our method in terms of its design choices, scalability, generalizability,
  and comparisons to baselines, including the most recent RL-based TAP solu
 tion. We also contribute the first benchmark for TAP which covers a variet
 y of input settings and difficulty levels.\n\nRegistration Category: Full 
 Access\n\nSession Chair: Chi Wing Fu (The Chinese University of Hong Kong)
URL:https://asia.siggraph.org/2023/full-program?id=papers_419&sess=sess157
END:VEVENT
END:VCALENDAR