Research

The Spatiotemporal Data Analysis (SDA) research group does research on machine learning methods for forming and analyzing spatiotemporal data to advance sustainability science. Research aims at methodological advances in machine learning, including uncertainty-aware learning, flow matching, representation learning, and multi-objective (hierarchical) reinforcement learning. These methods are designed to address the challenges of data efficiency, generalization, and interpretability, especially in high-stakes domains such as GNSS resilience, vision and sensor systems, and sustainable urban planning.
Sustainable smart cities by reinforcement learning algorithm development

In the fight against climate change and coping with the pressures arising from urbanisation, cities worldwide make solutions modifying pollutant emissions and thermal comfort. At present, road transport contributes a significant amount to the total carbon dioxide emissions in the EU. Thus, cities look for practical strategies to make their transport system more efficient and sustainable, and SDA Group is developing novel reinforcement learning methods to aid in this goal while maximising the air quality and minimising the carbon dioxide emissions.

Selected publications

  1. Matti Leinonen, Ahmed Al-Tachmeesschi, Banu Turkmen, Nahid Atashi, Laura Ruotsalainen (2024). Long Short Term Memory Based Traffic Prediction Using Multi-Source Data. International Journal of Intelligent Transportation Systems Research .
  2. Joanne C. Demmler, Ákos Gosztonyi, Yaxing Du, Matti Leinonen, Laura Ruotsalainen, Leena Järvi, Sanna Ala-Mantila (2021): PLOS ONE,
  3. Klavdiia Bochenina, Anton Taleiko, Laura Ruotsalainen: Simulation-based origin-destination matrix reduction: a case study of Helsinki city area, SUMO User Conference 2023.

Projects

  1. . This project was funded by the Research Council of Finland for 2020-2024.
  2. AI-based optimisation tool for sustainable urban planning (AIOSut). This project is funded by the Research Council of Finland for 2025-2026.
  3. . This project has received funding from the European Union – NextGenerationEU instrument and was funded by the Research Council of Finland for 2022-2024.
Representation learning for high dimensional sensor systems

The goal of our research is to develop advanced methods in representation learning and neural ordinary differential equations (ODEs) to model and understand dynamic, high-dimensional real-world sensor systems. These systems, such as those found in autonomous vehicles, industrial infrastructure, or environmental monitoring networks, generate complex spatiotemporal data that evolve continuously over time. By combining structured representation learning with data-driven differential modeling, we aim to extract compact, interpretable, and predictive representations that capture the underlying dynamics of these systems, enabling more robust decision-making, forecasting, anomaly detection and failure prediction in real-world conditions.

 

Selected publications

  1. Sarapisto, T., Wei, H., Heljanko, K., Klami, A., Ruotsalainen, L. (2024). Subsystem Discovery in High-Dimensional Time-Series Using Masked Autoencoders. European Conference of Artificial Intelligence.
  2. Farahani, A. V., Leinonen, M., Ruotsalainen, L., Jokisalo, J., & Kosonen, R. (2025). Predicting summer indoor temperatures in Nordic apartments considering heatwaves forecasts. Energy and Buildings, 115630.
GNSS signal protection and sensor data understanding

Availability of a global navigation and timing solution provided by GNSS has accelerated the appearance of various applications supporting sustainable development in the forms of improved safety, decreased environmental burden, new business opportunities and increased quality of life. However, the applications have demands on the PVT reliability, availability and accuracy that GNSS was not developed to support. The need for an ubiquitous GNSS that can adapt to the signal environment is self-evident. The development of accurate deep learning (DL) methods has provided huge improvements in computation in many disciplines. However, GNSS and navigation field has been slow to introduce DL into research. SDA group provides novel machine learning methods for improving the performance of a GNSS receiver and providing ubiquitous adaptive GNSS.

Selected publications

  1. Yan, Z., & Ruotsalainen, L. (2025). GNSS jammer localization in urban areas based on prediction/optimization and ray-tracing. GPS Solutions, 29(1), 47.
  2. Mehr, I. E., Savolainen, O. M. M., Ruotsalainen, L., & Dovis, F. (2024). Dual-Stage Deep Learning Approach for Efficient Interference Detection and Classification in GNSS. In Proceedings of the 37th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2024) (pp. 3336-3347). Institute of Navigation.
  3. Savolainen, O., Elango, A., Morrison, A., Sokolova, N., & Ruotsalainen, L. (2024). GNSS Anomaly Detection with Complex-Valued LSTM Networks. In J. Nurmi, R. Berkvens, T. Janssen, R. Halili, E. De Poorter, & A. Ometov (Eds.), 2024 International Conference on Localization and GNSS, ICL-GNSS 2024 - Proceedings Institute of Electrical and Electronics Engineers Inc..
  4. Savolainen, O., Malmivirta, T., Yan, Z., Morrison, A., & Ruotsalainen, L. (2024). Towards Jammer Fingerprinting: The Effect of the Environment and the Receiver to a Jammer Classification. In J. Nurmi, R. Berkvens, T. Janssen, R. Halili, E. De Poorter, & A. Ometov (Eds.), 2024 International Conference on Localization and GNSS (ICL-GNSS) Institute of Electrical and Electronics Engineers Inc..
  5. Arul Elango, Sahar Ujan, Laura Ruotsalainen (2022):
    2022 International Conference on Localization and GNSS (ICL-GNSS),
  6. Arul Elango, Ahmed Al-Tahmeesschi, Mikko Saukkoriipi, Titti Malmivirta, Laura Ruotsalainen (2022): 
  7. Ahmed Al-Tahmeesschi, Jukka Talvitie, Miguel López–Benítez, Laura Ruotsalainen (2022): 2022 International Conference on Localization and GNSS (ICL-GNSS), 

Projects

Master's theses

  1.  (2022).

News

  1.  (9 Jan 2024, in Finnish). 
Computer vision and sensor fusion

Computer vision provides crucial information for various applications requiring reliable navigation and situational awareness. SDA group develops novel methods for Simultaneous Localization and Mapping (SLAM) in challenging environments and 3D object tracking for improved safety for autonomous systems.

Publications

  1. Pyykölä, S., Joswig, N., & Ruotsalainen, L. (2024). Non-Lambertian surfaces and their challenges for visual SLAM. IEEE open journal of the Computer Society, 5, 430 - 445.
  2. Pajula, I. A., Joswig, N., Morrison, A., Sokolova, N., & Ruotsalainen, L. (2023). A Novel Cross-Attention-Based Pedestrian Visual–Inertial Odometry With Analyses Demonstrating Challenges in Dense Optical Flow. IEEE journal of indoor and seamless positioning and navigation, 2.
  3. Niclas Joswig, Juuso Autiosalo, Laura Ruotsalainen (2023): Machine Vision and Applications, 
  4. Laura Ruotsalainen, Aiden Morrison, Maija Mäkelä, Jesperi Rantanen, Nadezda Sokolova (2021): IEEE Sensors Journal,

Projects

Mas­ter's theses

  1.  (2021).
  2.  (2021)
  3.  (2022).

News

  1.  (15.1.2020, in Finnish)

Videos