BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240626T180033Z
LOCATION:3001\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240625T113000
DTEND;TZID=America/Los_Angeles:20240625T114500
UID:dac_DAC 2024_sess103_RESEARCH883@linklings.com
SUMMARY:RT-MDM: Real-Time Scheduling Framework for Multi-DNN on MCU Using 
 External Memory
DESCRIPTION:Research Manuscript\n\nSukmin Kang, Seongtae Lee, and Hyunwoo 
 Koo (Sungkyunkwan University); Hoon Sung Chwa (Daegu Gyeongbuk Institute o
 f Science and Technology); and Jinkyu Lee (Sungkyunkwan University)\n\nAs 
 the application scope of DNNs executed on microcontroller units (MCUs) ext
 ends to time-critical systems, it becomes important to ensure timing guara
 ntees for increasing demand of DNN inferences. To this end, this paper pro
 poses RT-MDM, the first real-time scheduling framework for multiple DNN ta
 sks executed on an MCU using external memory. Identifying execution-order 
 dependencies among segmented DNN models and memory requirements for parall
 el execution subject to the dependencies, we propose (i) a segment-group-b
 ased memory management policy that achieves isolated memory usage within a
  segment group and sharded memory usage across different segment groups, a
 nd (ii) an intra-task scheduler specialized for the proposed policy. Imple
 menting RT-MDM on an actual system and optimizing its parameters for DNN s
 egmentation and segment-group mapping, we demonstrate the effectiveness of
  RT-MDM in accommodating more DNN tasks while providing their timing guara
 ntees.\n\nTopic: AI\n\nKeyword: AI/ML Algorithms\n\nSession Chairs: Hongya
 ng Jia (Tsinghua University) and Grace Li Zhang (Technische Universität Da
 rmstadt)
END:VEVENT
END:VCALENDAR
