BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:ENAC Seminar Series by Dr C. Andriotis
DTSTART:20190923T130000
DTEND:20190923T140000
DTSTAMP:20260503T220621Z
UID:d22195900675d5229b3f2896a6d153e6c052736caafef98e048e9c77
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr Charalampos Andriotis\n13:00 – 14:00 – Dr Charalampos A
 ndriotis\nResearch Assistant\, Pennsylvania State University\, USA\n\nAI-d
 riven decision-making for intelligent infrastructure: enabling the built e
 nvironment to reason and adapt\n\nOver the past 5-10 years\, we have witne
 ssed groundbreaking advances in data analytics and Artificial Intelligence
  (AI)\, which have prompted transformative changes in society\, science an
 d engineering. These advances come with tremendous opportunities for civil
  engineering to expand its impact to the prospects of the fourth industria
 l revolution\, and address the unprecedented challenges posed by an aging\
 , growing and changing built environment. In view of multiple hazards\, in
 creased population demands\, climate change effects\, resource limitations
 \, and big data delivering incomplete information\, intelligent infrastruc
 ture management and design requirements exceed the limits of conventional 
 approaches\, necessitating innovative data-driven decision-making and info
 rmatics frameworks\, able to support nimble and robust mitigation of the i
 nvolved socioeconomic and environmental risks.\n\nSuch frameworks need to 
 efficiently (i) facilitate learning and reasoning from real-time\, noisy o
 bservational and/or sensory data\; (ii) quantify the value-of-information 
 of monitoring technologies towards improved decision-making\; (iii) inform
  detailed decision plans in complex multi-component networks operating in 
 high-dimensional stochastic domains\; (iv) ensure long-term optimality and
  adaptability over the entire system life-cycle\; and (v) be robust to res
 ource and/or sustainability-related constraints. Bridging AI and engineeri
 ng decision-making\, the development of novel methods that address these n
 eeds is discussed in this talk\, within the contexts of stochastic dynamic
  programming\, dynamic Bayesian networks\, partially observable Markov dec
 ision processes\, deep learning and centralized/decentralized multi-agent 
 reinforcement learning.\n\nThe developed approaches target\, but are not l
 imited to\, structural applications and civil infrastructure management\, 
 and are analyzed as per their remarkable strengths in providing solutions 
 to otherwise intractable decision problems. The contributions and future d
 irections of the developed line of research are presented\, along with the
 ir capabilities in enabling decision-making in high-dimensional state/acti
 on spaces\, value-of-information assessment\, life-cycle optimality\, real
 -time autonomous strategy adaptation in view of new information\, flexible
  policies that effectively and timely avert future risks\, and resilient r
 ecovery from extreme event catastrophes.\n 
LOCATION:INJ 218 https://plan.epfl.ch/?room==INJ%20218
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR