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:20260402T024533Z
LOCATION:3003\, Level 3
DTSTART;TZID=America/Los_Angeles:20250624T164500
DTEND;TZID=America/Los_Angeles:20250624T170000
UID:dac_DAC 2025_sess123_RESEARCH1390@linklings.com
SUMMARY:Towards Uncertainty-aware Robotic Perception via Mixed-signal BNN 
 Engine Leveraging Probabilistic Quantum Tunneling
DESCRIPTION:Likai Pei, Yu Zhou, Xingtian Wang, Xueji Zhao, Wanxin Huang, a
 nd Boyang Cheng (University of Notre Dame); Halid Mulaosmanovic, Stefan Du
 enkel, Dominik Kleimaier, and Sven Beyer (GlobalFoundries); and Kai Ni, Me
 ngxue Hou, Michael Niemier, and Ningyuan Cao (University of Notre Dame)\n\
 nIntegrating deep learning with environmental perception enhances robotic 
 adaptability to complex tasks. However, its ``black-box'' nature, such as 
 the lack of uncertainty quantification, poses challenges for safety-critic
 al applications, particularly in unstructured and noisy environments. Baye
 sian neural networks (BNNs) offer uncertainty quantification but are limit
 ed by high hardware overhead, restricting real-time implementation on reso
 urce-constrained robots. This paper presents a mixed-signal hardware accel
 erator for BNNs, utilizing probabilistic quantum tunneling in fully deplet
 ed silicon-on-insulator (FD-SOI) transistors to enable efficient, real-tim
 e uncertainty quantification. Device measurements indicate high-quality Ga
 ussian random variable generation, validated through quantile-quantile plo
 t analysis, with a high correlation coefficient (r = 0.997) at 200 fJ/samp
 le. Leveraging such compact randomness, the parallel architecture achieved
  10^3--10^4× latency reduction at less than 2 × area cost. Finally, in unc
 ertainty-aware visual localization application of autonomous underwater ve
 hicles, the BNN model effectively distinguishes data noise from model unce
 rtainty, yielding significant information gain and enhancing the resamplin
 g efficiency by  4.5× at same accuracy.\n\nTopics: Design\n\nTracks: DES4:
  Digital and Analog Circuits\n\nSession Chairs: Ioannis Savidis (Drexel Un
 iversity) and Kishor Kunal (Nvidia)\n\n
END:VEVENT
END:VCALENDAR