International Projects

Other Collaborations

Peggy Guggenheim Collection -- VAPOUR

VAPOUR (Visitor Analytics for Pollution, Occupancy, Usage, and Routes) is a research project between CNR-IAC and the Solomon R. Guggenheim Foundation. The project aims to monitor, analyse, and optimise visitor dynamics within crowded museum, while studying the environmental impact of visitors flow on indoor air quality and conservation conditions. Peggy Guggenheim Collection in Venice, Italy, is adopted as a case study due to the peculiar environmental conditions the city of Venice provides.

Combining Eulerian and Lagrangian measurement strategies, the project relies on both fixed and mobile tracking tools: a custom IAC-developed mobile app for room entry counts, a Bluetooth-based infrastructure (developed in CURATOR project) to collect high-resolution visitor trajectories using beacons and Raspberry Pi receivers, and different commercial tools for pollutant and hygrometric data gathering.

VAPOUR project enables the accurate reconstruction of visitor routes, occupancy levels, and stopping times in front of individual artworks. These behavioural insights are then cross-analysed in collaboration with CNR-ISPC with measurements of CO2, humidity, and pollutant levels to assess microclimatic stress induced by crowding.

Key outcomes include:

• quantitative assessment of room congestion during permanent and temporary exhibitions

• identification of high-attraction artworks through “holding power” analysis

• museum layout optimisation proposals, which have influenced the renovation planning in 2024

• ongoing studies on the correlation between environmental parameters and visitor presence

Autostrade Alto Adriatico -- SPARTA 2.0 & SPARTA 3.0

SPARTA (Smart Predictive Analysis for Road Traffic Applications) is a research collaboration between CNR-IAC and Autostrade Alto Adriatico S.p.A. (formerly Autovie Venete S.p.A.), headquartered in Palmanova, Italy. The projects focus on the analysis and prediction of vehicular traffic, combining data-driven techniques and macroscopic traffic flow models. The regional highway network of Friuli and Venezia Giulia are used as a case study and proof of work.

In SPARTA 2.0, we developed a hybrid approach that integrates Long Short-Term Memory (LSTM) neural networks with classical first-order macroscopic traffic models (LWR). Using real highway data from fixed sensors—classified by lane and vehicle type—we employed LSTM models to predict both the short-term evolution of traffic flux and the onset of congestion. These machine learning outputs were then incorporated into the differential model in two key ways: by inverting the fundamental diagram to estimate vehicle density from flux data, and by setting more accurate boundary conditions for the simulation. This hybrid strategy improves the reconstruction of traffic dynamics, especially under non-recurrent events such as accidents in poorly instrumented road segments.
> [DOI:10.1007/s10444-024-10206-8]

In SPARTA 3.0, the focus shifted to unsupervised learning and anomaly detection. By applying clustering techniques (hierarchical and partitioning methods) to time-series representations of traffic data, we developed tools for the automatic identification of anomalous patterns, including sensor malfunctions and irregular congestion episodes. The methodology enables real-time detection of deviations from expected traffic behaviour, providing a practical tool for infrastructure operators. In particular, symbolic representations and Dynamic Time Warping were explored to enhance the robustness and interpretability of the clustering results.
> [arXiv:2504.00881]

GALLERIA BORGHESE -- CURATOR

CURATOR (Crowd Understanding and Real-time Analytics for Traffic Optimisation and Routing) is a research initiative aimed at the analysis, prediction, control, and optimisation of visitors’ movement within crowded museums. The project merges artificial intelligence, stochastic modelling, and IoT technologies to create a comprehensive digital twin of human flow in indoor cultural spaces. The ultimate goal is to improve visitor experience, safety, and logistical efficiency, while supporting the preservation of delicate artworks.  The Galleria Borghese museum in Rome is adopted as a case study.

At the core of CURATOR is a non-invasive Bluetooth-based localisation system, where visitors are equipped with lightweight beacons tracked by Raspberry Pi sensors discreetly placed in the museum. By applying advanced filtering algorithms—including a sliding window statistical model and a neural network-based RSSi regressor—we reconstruct individual room-scale trajectories with high accuracy (>96%), even under low sensor density.

The recorded paths are then analysed through clustering techniques and custom trajectory-space metrics, yielding a probabilistic transition model (time varying Markov model) between rooms.  The TVMM is used to generate synthetic visitor dynamics and to simulate different crowding scenarios. The results directly support data-driven ticketing strategies and crowd flow optimisation, allowing the museum to safely increase capacity while preserving comfort and artistic integrity.

> [DOI:10.1016/j.jocs.2021.101357][Link to other publications]