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:20211208T164413Z
LOCATION:3016
DTSTART;TZID=America/Los_Angeles:20211209T143000
DTEND;TZID=America/Los_Angeles:20211209T145000
UID:dac_DAC 2021_sess102_RESEARCH1012@linklings.com
SUMMARY:Analyzing and Improving Fault Tolerance of Learning-Based Navigati
 on Systems
DESCRIPTION:Research Manuscript\n\nAnalyzing and Improving Fault Tolerance
  of Learning-Based Navigation Systems\n\nWan\n\nLearning-based navigation 
 systems are widely used in autonomous applications, e.g., robotics, unmann
 ed vehicles and drones. Specialized hardware accelerators have been propos
 ed for high-performance and energy-efficiency for such navigational tasks.
  However, transient and permanent faults are increasing in hardware system
 s and can catastrophically violate tasks safety. Meanwhile, traditional re
 dundancy-based protection methods are challenging to deploy on resource-co
 nstrained edge applications. In this paper,we experimentally evaluate the 
 resilience of navigation systems regarding algorithms, fault models and da
 tatypes from both RL training and inference. We further propose two cost-e
 fficient fault mitigation techniques that achieve 2x success-rate and 39% 
 quality-of-flight improvements in learning-based navigation systems\n\nTop
 ic: Autonomous Systems\n\nVirtual Program: Presented In-Person
END:VEVENT
END:VCALENDAR