Informatics Colloquium 2. 5. Why should the authorities lie to you
Informatics Colloquium 2. 5. 2017, 14:00 lecture hall D2
Dr. Jakub Mareček, IBM Ireland Research Laboratory
Why should the authorities lie to you
Abstract: Many real-world problems in Transportation Engineering are, in effect,
control problems with rather limited identifiability of the underlying
non-linear problems. Consider, for instance, the problems of information
provision: an authority measures the travel times across the network, announces
the travel times to the public, who base their route choices on the announced
travel times, and are measured, eventually. The relationship between the number
of concurrent users of a road-segment and the time they spend traveling across
the segment is non-linear; the route choice is not deterministic and one cannot
"perfectly excite the system", as in lie to the public, without the public
noticing and changing their behaviour. At the same time, it is clear that many
approaches currently in use are failing: consider two parallel routes and the
authority announcing two distinct travel times (historical, current, or forecast
by any method whatsover). Most drivers pick the route announced as faster,
thereby congesting it (and invalidating any forecast). Similar issues arise in
traffic control systems, which control traffic lights at junctions. In a series
of papers, we have developed an approach to closed-loop analyses of such systems
and we will present some of the key insights.
Short biography: Jakub Marecek is a research staff member at IBM Research.
Together with some fabulous colleagues, Jakub develops solvers for optimisation
and control problems in IBM's Smarter Cities Technology Centre. His recent work
includes polynomial optimisation in power systems, policies for bi-level
optimisation, and a stream processing system for urban traffic management
(called "Insight"), which has just won the 5th Annual Award for Excellence by
ITS Ireland. Jakub is also the principal investigator for VaVeL, an H2020
project within the "Big Data" call, and a programme committee member for AISTATS
2017 and ICAPS 2017. Prior to joining IBM in August 2012, Jakub had worked on
distributed solvers for non-smooth convex optimisation problems at the
University of Edinburgh and on integer programming at the University of
Nottingham. Jakub grew up in Brno, the Czech Republic, where he had worked in
two start-up companies before studying for his first two degrees at Masaryk
University.