The primary objective of this consortium project performed in 2021-2024 was to develop methods and tools focusing on understanding urban mobility and promoting the use of environmentally friendly travel modes. Our group in the CoMobility project was responsible for the development of ​​machine learning methods and large-scale data analysis with Big Data platforms and newly developed software modules, focusing on modelling travel mode choices with machine learning methods. This relied on novel survey data collected under the supervision of the SGH Warsaw School of Economics documenting trips and preferences of the citizens of Warsaw, Poland.

Project partners: University of Warsaw (leader), The Norwegian Institute for Air Research (Norway), Warsaw University of Technology (WUT), The SGH Warsaw School of Economics, The Fridtjof Nansen Foundation at Polhøgda (Norway), The On-site Foundation (Poland), City of Lublin (Poland)

Funding: Icelend, Liechtenstein, Norway Grants
Call: IdeaLab: Cities for the future: services and solutions
Grant agreement No: NOR/IdeaLab/CoMobility/0001/2020-00
Project manager at the Warsaw University of Technology: Maciej Grzenda

In the CoMobility project, we focused on predicting the use of different transport modes with machine learning methods and understanding under what conditions citizens decide to use environmentally-friendly modes. First of all, we developed tools for processing large-scale data such as GPS traces of up to 2,000 public transport vehicles operated at the same time by the City of Warsaw. We have collected real-time data on vehicle locations and daily schedules for over 600 days. This provided basis for identifying planned and real public transport graphs, identifying planned and actually feasible connections and quantifying them with level of service features. To recreate other conditions influencing travel behaviours, we have collected and processed also weather and air pollution time series, and spatial data of built and natural environment and combined it with demand data from transport model. This enabled major improvements in the prediction of travel mode choices, which we performed with batch and stream mining methods.

The aim of this project performed in 2020-2022 was to develop machine learning methods dedicated to data streams. The basic assumption was to take into account the actual conditions of forecasting with the use of such data, including missing labels, delays in the collection of measurement values, e.g. resulting from problems with wireless communication, delayed labels causing verification latency, e.g. delayed availability of information about the actual type of failure. An important aspect of the project was international cooperation, including the collaboration with researchers from LTCI, Télécom Paris, France and University of Waikato, Hamilton, New Zealand.

Project partners: Warsaw University of Technology

Funding: Excellence Initiative - Research University. Warsaw University of Technology
Call: POB Research Centre for Artificial Intelligence and Robotics/SzIR-1
Project manager at the Warsaw University of Technology: Maciej Grzenda

In the project, we developed inter alia a novel method of quantifying volatility of predictions of stream mining models and showed that stream mining models frequently change their predictions for the same input data, though these changes in many and event most cases do not reduce prediction errors. Moreover, we contributed to a survey of semi-supervised methods under delayed labelling setting on which we collaborated with researchers from inter alia University of Waikato and Institut Polytechnique de Paris. Furthermore, we developed a data imputation method for data streams and applied it to public transport data streams.

The VaVeL project performed in 2015-2018 aimed to largely improve urban data usage, focusing on transportation and urban challenges. By developing a framework to manage and analyse diverse, noisy data streams, it addressed inefficiencies in current technologies. The project was performed with two European cities, three companies, and expert researchers. VaVeL seeked to make urban data more accessible, benefiting industries and improving city life through impactful, real-world solutions.

Project partners: National and Kapodistrian University of Athens (the coordinator), Technische Universitat Dortmund, Technion – Israel Institute of technology, Fraunhofer Institute for Intelligent Analysis and Information Systems, IBM Ireland Limited, AGT Group (R&D) GmbH, Orange Polska S.A., Warsaw University of Technology, Dublin City Council, City of Warsaw