Firms are increasingly investing in AI to support their operational decision-making processes. Together with On Track Lab and VAIA - Flanders AI Academy we want to create a realistic picture of the opportunities and limitations of AI for decision support via real-time dashboards via this workshop. A special focus will be given to control rooms as this sector is increasing in importance with the digitization of operations. The workshop covers three topics:
FORMAT Study day/workshop
TIMING October 7, 2022
LOCATION University Ghent, Faculty of Economics and Business Administration, Faculty board room/Faculteitsraadzaal
Industry: 250 €
Researchers: 80 €
Technical skill level: Basic knowledge of AI terminology is a prerequisite (you know what AI is), no mathematical/coding experience is required.
Job profiles: Digital managers, data scientists and researchers in the field of digital operations, and predictive analytics.
Industries: Safety-critical settings, utilities, transportation, control rooms.
9.00 Coffee reception
9.40 Railway traffic control room, and room for control
prof. dr. Francesco Corman
11.00 coffee break
11.20 Developing Decision support systems for Traffic Control
12.40 lunch + (small) poster session
13.40 Democratizing business intelligence and machine learning for air traffic management safety
Prof. Dr. Riccardo Patriarca, Sapienza Universitá di Roma and Eurocontrol
14.20 Artimation: Transparent Artificial intelligence and Automation to Air Traffic Management Systems
15.00 coffee break
15.20 Putting the AI in air traffic control
Evelina Gabasova, The Alan Turing Institute
16.00 Debate: Industry vision and expectations of AI in the control room
Peter Prater, Founder & Chair of the International Critical Control Rooms Alliance (ICCRA)
16.40 Closing workshop & networking drink
Marijn Verschelde, IÉSEG School of Management
We review different challenges and opportunities for traffic control in railway systems. From the point of view of sensing, state estimation and data fusion must be performed in a very short time, and for units that are typically spatially dispersed. From the point of view of determining a control objective to support automatic decisions, the challenge is how to understand the impact of a decision in terms of system performance. Almost all of those problems have to deal with unknown future states, which must be predicted, typically by model-based or black box approaches, also based on advanced analytics. Once an objective function and optimization variables are determined, optimization models can help to find a solution quickly and effectively. Further challenges are the acceptance of decision stakeholders, within the control room, but also within the travelers and operators, or the direct implementation of automatic digital control. For passenger-oriented traffic control, this is particularly interesting and challenging, due to the large amount of possible decisions per decision maker, and data that can partially describe those aspects, which calls for machine learning approaches. To reach all those goals, the system must have room for control, in another sense, flexibility in operations must be built in already from the planning, to be exploited in the real-time horizon when needed.
by Christophe Hurter - ENAC
The Decision Making Process is already associated with AI. The algorithms are meant to help ATCOs in daily tasks, but they still face acceptability issues. Today’s automation systems with AI/Machine Learning do not provide additional information on top of the Data Processing result to support its explanation, making them not transparent enough. The Decision Making Process is expected to become a “White Box”, giving understandable outcome through an understandable process. XAI SOLUTIONS: Transparency and Explainability: ARTIMATION’s goal is providing a transparent and explainable AI model through visualization, data driven storytelling and immersive analytics. This project will take advantage of human perceptual capabilities to better understand AI algorithm with appropriated data visualization as a support for explainable AI, exploring in the ATM field the use of immersive analytics to display information.
by Riccardo Patriarca - Sapienza Université di Roma & Radu Ciponea - Eurocontrol
The ways in which Air Navigation Service Providers (ANSPs) monitor safety performance is strongly influenced by international regulations, standards, and agreements, although each State may also add its own local requirements. Particularly in the case of more mature ANSPs, the regulatory safety performance obligations are merely the tip of the iceberg in the undertaken safety performance activities. Much of the indicators, methods and tools are over and above what is required by regulations, either national or international. In modern settings, the usage of Business Intelligence and Machine Learning solutions can be enumerated under the continuous chasing of strategies to foster ANSPs’ safety intelligence capacities towards higher standards.
This manuscript shows the development process of an integrated data-driven framework for self-service BI and ML on safety reporting data for the air traffic management system. The proposed framework firstly focuses on the development process of a BI architecture to extract meaningful knowledge from multiple data sources. Then, it progresses discussing how ML solutions may support gaining a deeper understanding of system’s performance and delineating specific safety recommendations. The explorative application of the proposed framework in multiple European ANSPs provides the basis for sharing lessons learned and outlining a possible path to start democratizing safety intelligence in aviation.
Digitisation and employee workload (im)balance are intertwined. To address undesirable workload peaks and lows, we propose a 2-step machine learning model to provide real-time workload analytics per controller in digital safety-critical control rooms. The advocated model leverages a rich real-time data structure with disaggregated event-level taskload data. Next to exploring different machine and deep learning approaches, we compare the performance of a model that predicts aggregate workload with the performance of the aggregate of different models that predict specific task loads. We develop a business application that utilizes the proposed model to provide detailed predictive analytics that open the black box of workload imbalance and, in this way, empowers the control room manager with real-time insights.
Dr Evelina Gabašová is a data scientist and machine learning researcher, working in The Alan Turing Institute, the UK’s national centre for data science and artificial intelligence. She writes about data science, machine learning and software development.
Christophe Hurter is a Professor working at the University of Toulouse, France, leading the Interactive Data Visualization group (DataVis) of the French Civil Aviation University (ENAC). His research covers explainable A.I. (XAI), big data manipulation and visualization (InfoVis), immersive analytics, and humancomputer interaction (HCI). He investigates the design of scalable visual interfaces and the development of pixel-based techniques. He is an associate researcher at the research center for the French Military Air Force Test Center (CReA, Base militaire de Salon de Provence) and at the Brain and Cognition Research Center (CerCo, Hospital University Center of Toulouse). He published 2 books, 4 book chapters, 20 patents, 25 journal papers, more than 100 per reviewed international research papers.
Riccardo Patriarca is a tenure track assistant professor at Sapienza University of Rome (Italy) – Dept. of Mechanical and Aerospace Engineering. He holds an BSc in Aerospace Engineering, an MSc in Aeronautical Engineering and a PhD in Industrial and Management Engineering (Doctor Europaeus). He has published widely (about 100 manuscripts published in academic journals and conference proceedings) on methodological and epistemological aspects of risk, safety, and resilience management as well as operations management in general. He aims to make systems safer and resilient when - and especially before - things go awry.
Bart is a Belgian railroader. He has extensive managerial and expert experience at Belgian railways, and international lobbying experience at the Community of European Railways (CER). Bart co-founded the On Track Lab, and currently serves as a Principal Engineer at Infrabel’s Performance Data division. His primary goal is to bridge the gap between practitioners and academics, and foster result-driven collaboration that is beneficial to both parties.
He has developed partnerships with academics from Belgium, France, the UK, the USA, and Australia. Bart holds a MSc in Electronics Engineering, a MSc in Public Management, and a PhD in Business Economics (all from Ghent University, Belgium). He is affiliated researcher at Ghent University, and adjunct professor at IÉSEG School of Management (France). He is also partner and key collaborator in several research projects at Virginia Tech’s System Performance Lab (USA).
Léon Sobrie is a Ph.D. Student at Ghent University – Faculty of Economics and Business Administration. He holds a BSc and MSc in Business Engineering. His research focuses on developing and implementing machine learning models for predicting key metrics (delays, workload, safety) in digital control rooms aiming to provide real-time analytics for managers. These research efforts are executed in the context of the On Track Lab, a multi-disciplinary research lab on operations, transportation and network analytics founded by Ghent University, Infrabel and IÉSEG School of Management. His work, co-authored by Marijn Verschelde and Bart Roets, has been presented at international conferences (INFORMS 2021 (Anaheim, US ) & EURO2021 (Athens, Greece)).
Marijn Verschelde is associate professor in quantitative methods at IÉSEG School of Management (LEM-CNRS 9221, Lille, France). He is also visiting professor at KU Leuven and co-founder of the On Track Lab. Further, he is co-coordinator of the new LEM research group Operations Research & Organizations Performance (OPER). His proven track record includes next to publications in the field of operations and applied micro-econometrics, collaborations with companies and central banks concerning firm performance. His most recent research focuses on machine learning for decision making.
Hakan Ergun, born in 16.02.1983 in Leoben Austria, has obtained his degree of Master of Science in Electrical Engineering at the Graz University of Technology (TU Graz) in October 2009. In February 2010 he has joined the Electa Research Group at KU Leuven, Belgium where he obtained his PhD in Electrical Engineering in January 2015. He has been a post – doctoral researcher until 2018 and is currently a research expert at KU Leuven / EnergyVille.
His main research interests are optimization methods in power system planning, power system modelling, power system reliability, electricity markets and regulation. To this date he has worked in several national and international research projects on several aspects of power system modelling and optimization.
He has published several papers in international scientific journals and conferences. He is a senior member of IEEE and is an active member of CIGRE. He has been in the local organization committee of the IEEE EnergyCon 2016 conference in Leuven. He is a vice chair of the IEEE PES/PELS/IAS Benelux Chapter and has been the chapter chair between 2017 and 2018. He is an associate editor with the Canadian Journal of Electrical and Computer Engineering.
Hussain Kazmi is currently an FWO postdoctoral research fellow at KU Leuven, where his research is at the intersection of machine learning, optimal decision making and energy. His core area of expertise lies in developing algorithms that integrate domain expertise and downstream task information into data-driven models for smart(er) energy systems. Very recently, this research work has received the annual award of the International Institute of Forecasters. Currently, he leads a cross-European EIT InnoEnergy working group aimed at developing a data science program for energy engineers.
He holds a PhD at the intersection of data science and energy engineering from KU Leuven (2019), as well as MSc degrees in Sustainable Energy Technology, and Energy and Nuclear Engineering from Technical University of Eindhoven (The Netherlands) and Politecnico di Torino (Italy) respectively. In 2021, he was a visiting research scholar at KTH Royal Institute of Technology (Sweden). As the first data scientist at two different Belgian clean energy startups (Enervalis and iLECO), he has also helped set up data science teams in applied settings.
Peter Prater is a well-known face in the Public Safety ICT world, his commitment and passion evidenced through being the Founder and Chair of the International Critical Control Rooms Alliance (www.iccraonline.com), a long-time supporter of the TCCA and a Life Member of British APCO where he fulfilled many roles between 1994 and 2018.
Peter is Hexagon’s UK Managing Director for its Safety, Infrastructure & Geospatial division. Since taking up this appointment the team secured the contract to replace London’s Metropolitan Police Service’s aged command and control solution. He is passionate about the role and operation of critical control rooms and particularly how these are influenced by changes in technology and how they can support the Safe Cities agenda
Prior to Hexagon, Peter started his working life as a user when he served for 15 years throughout the world with the British Army’s Royal Corps of Signals. Following his army career Peter worked as an independent consultant for 14 years, rising to Head of ICT Consulting at Hyder Consulting before joining Frequentis in 2009 in the role of Key Account Manager for its relationship with the Metropolitan Police Service. Peter’s background then is firmly based in the deployment and use of mobile and fixed communications and command and control systems in support of critical operations.
Francesco Corman received the M.S. degree in management engineering from Roma Tre University in 2006 and the Ph.D. degree from the Delft University of Technology in 2010. He is currently the Chair of the Transport Systems (Assistant Professor) at the Swiss Federal Institute of Technology, ETH Zurich, with main responsibilities in research and education in transport systems with a particular focus on analytics and optimization methods for railways, public transport and logistics system and their interconnection, with special focus on their operations.
Wilco Tielman is a data scientist at ProRail since 2015 and a guest researcher at the University of Utrecht within the Intelligent Systems group. He got his Msc degree in Technical Artificial Intelligence at the University of Utrecht. He has been working at ProRail on data science and AI related questions for the traffic control department, with the aim of closing the gap between research and practice.