ACROSS–DataScience

Znanstveni centar izvrsnosti za znanost o podatcima i kooperativne sustave

Centar ACROSS-DataScience djeluje kao središte suradnje akademske zajednice s poslovnim i javnim sektorom u istraživanju i razvoju novih metodologija i naprednih inženjerskih pristupa za znanost o podatcima i kooperativne sustave.


Poziv na predavanje:...

Treće predavanja u okviru istraživačkog seminara pod naslovom "Application of Machine Learning for Traffic Control on Urban Motorways" će u petak 22. rujna u 13:00 sati u dvorani D1, O69, ZUK Borongaj održati Martin Gregurić, mag. ing. traff. Životopis predavača te kratki opis predavanja možete pročitati u opširnijem sadržaju obavijesti.

Abstract: To tackle today’s problems with congestions in road traffic and the inability to expand roads in an urban environment, new solutions in the form of advanced control methods on the existing road infrastructure are applied. Such solutions are from the domain of intelligent transportation systems (ITS). ITS essentially integrates information and communication technologies in order to resolve mentioned congestion problems. Often used ITS based traffic control methods on urban motorways are ramp metering (RM) and variable speed limit control (VSLC). They use traffic sensors to measure traffic parameters based on which a wider picture of current traffic situation is considered. A dedicated algorithm for RM or VSLC uses those sensed data to compute actions that will make a positive impact on traffic flows. During this talk, special emphasis will be set on two types of machine learning techniques that can be used in RM and VSLC algorithm design. Integration of an adaptive neuro-fuzzy inference system (ANFIS) as the control part and a recurrent neural network for traffic demand prediction will be applied for RM. To learn an ANFIS network first an appropriate learning data set has to be created using standard control approaches. Reinforced learning as an on-line learning method will be applied for VSLC. Results of both approaches will be discussed with a focus on the design of a suitable learning dataset and needed iterations.

Biography: Martin Gregurić received his Bachelor and Master Degrees in traffic and transport engineering, course: Intelligent transportation systems, in 2008. and 2011, respectively. In November 2011 he enrolled in the Ph.D. study Technological systems in traffic and transport at the Faculty of Transport and Traffic Sciences University of Zagreb. From July 2013 to September 2015 he was employed as a research assistant in the project: Intelligent Cooperative Sensing for Improved Traffic Efficiency, FP7-317671 ICSI. He is partly involved in lectures of courses: Automatic control in traffic and transport on the undergraduate level, and Intelligent transportation systems I and II, Virtual reality in traffic, and Artificial Intelligence at the graduate level. He received the University of Zagreb Rector award, award “Best graduate student in the course of Intelligent transportation systems in 2010”, and won the third prize at the Autonomic Road Transport Support Systems Early Career Researcher Conference (La Valletta, Malta, 2015). He participated in several international COST Training Schools with the main topic of autonomic road transport support systems, first summer school of the Croatian Centre of Research Excellence for Data Science and Advanced Cooperative Systems: Research Unit Data Science and the summer school “Intelligent Cars on Digital Roads – Frontiers in Machine Intelligence” organized by the BMW group.

Popis obavijesti