Pasar al contenido principal

DISCOVERY: PROTOCOLOS EN REDES DE COMUNICACIONES Y PRIVACIDAD DE DATOS

PID2023-148716OB-C32

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal para Impulsar la Investigación Científico-Técnica y su Transferencia
Subprograma Subprograma Estatal de Generación de Conocimiento
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2023
Plan Estatal 2021-2023 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSITAT POLITECNICA DE CATALUNYA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 30
Encontrada(s) 1 página(s)

UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

MobilitApp: a deep learning-based tool for transport mode detection to support sustainable urban mobility

  • Caravaca Ibáñez, Gerard
  • Cruz Llopis, Luis Javier de la|||0000-0003-4171-8310
  • Catalin Diaconeasa, Adrian
  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
The visible effects of climate change in urban areas are driving a paradigm shift in societal and political priorities. Urban planners, public transport providers, and traffic managers are increasingly focused on redesigning cities to promote sustainable mobility and create green spaces for pedestrians, cyclists, and scooter users. In alignment with these objectives, the European Climate Law mandates a minimum 55% reduction in greenhouse gas emissions by 2030 and climate neutrality by 2050. Achieving these targets requires robust tools to collect and analyze mobility data, enabling the evaluation of citizens’ travel habits and the planning of sustainable urban infrastructure. This study presents MobilitApp, a tool developed based on a deep learning (DL) model for real-time detection of transportation modes using smartphone sensor data. Our approach leverages a hierarchical model combining convolutional neural networks (CNNs) for feature extraction and long short-term memory (LSTM) layers for temporal processing, enhanced by skip connections. To ensure computational efficiency on mobile devices, the system integrates statistical techniques for early motion detection, minimizing reliance on DL models. The model was trained on a dataset of multimodal trips in Barcelona, achieving over 80% accuracy for most transport modes and a weighted average accuracy of 88%. These results highlight the effectiveness of our approach for accurately predicting users’ transport modes during their trips. The MobilitApp tool provides an intuitive platform for collecting and analyzing urban mobility data. By analyzing travel patterns, transport modes, and mode-switching behaviors, it delivers actionable insights to city planners, aiding in the enhancement of urban mobility, promotion of sustainable development, and transition to greener cities., This work was supported in part by Spanish Government funded by MCIN/AEI/10.13039/501100011033 under Research Projects ’’DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)’’ under Grant PID2023-148716OB-C32 and ’’Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)’’ under Grant PID2020-113795RB-C31; partially funded by MCIN/AEI/10.13039/501100011033 and by the European Union (EU) NextGenerationEU/PRTR (Plan de Recuperación, Transformación y Resiliencia) under Research Project ’’Anonymization Technology for AI-Based Analytics of Mobility Data (MOBILYTICS)’’ under Grant TED2021-129782B-I00; in part by the Predoctoral Scholarship ‘‘Generación de Conocimiento- Projects Call 2022’’ under Grant PRE2021-099830; and in part by the Generalitat de Catalunya under AGAUR(Agència de Gestió d’Ajuts Universitaris i de Recerca) Grant 2021-SGR-01413., Peer Reviewed, Objectius de Desenvolupament Sostenible::11 - Ciutats i Comunitats Sostenibles, Objectius de Desenvolupament Sostenible::11 - Ciutats i Comunitats Sostenibles::11.2 - Per a 2030, proporcionar accés a sistemes de transport segurs, assequibles, accessi­bles i sostenibles per a totes les persones, i millorar la seguretat viària, en particular mitjan­çant l’ampliació del transport públic, amb especial atenció a les necessitats de les persones en situació vulnerable, dones, nenes, nens, persones amb discapacitat i persones grans
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

An AI-based intelligent vehicle routing approach for large-scale fleet optimization

  • Guin, Agneev
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
This research addresses the challenges of urban mobility and logistics by developing an AI-based system for intelligent route prediction and coordination. Using Graph Neural Networks and complementary AI methods, we intend to model heterogeneous fleets to optimize large-scale mobility with real-time decision-making and energy-aware trade-offs., This work was supported by the DISCOVERY project PID2023-148716OBC31 funded by MCIU/AEI/10.13039/501100011033. Also, by the MultiMO project TSI-100123-2024-60. Finally, by the Generalitat de Catalunya under AGAUR grant 2021 SGR 01413., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Artificial intelligence methods for anomaly detection in energy consumption

  • Pérez Méndez, Ana Laura
  • Monteagudo Gordillo, René
  • Bazan Prieto, Carlos Alberto
  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Bello Pérez, Rafael E.
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
Non-technical losses represent a significant challenge for energy distribution companies due to their economic impact. These losses typically arise from irregularities at supply points and fraudulent customer behavior. In Cuba, electricity meter readings are performed manually, and consumption data is processed using spreadsheets that combine weighted criteria to generate alerts for potential anomalies. This procedure is prone to vulnerabilities such as manual data entry errors, incorrect key assignments, and human mistakes made by field readers, which compromise the reliability of the analysis. In this paper we present an anomaly detection system for electricity consumption, developed using artificial intelligence techniques to identify irregularities based on monthly reports from residential users. Various machine learning methods were evaluated, with eXtreme Gradient Boosting (XGBoost) standing out for its effectiveness in handling imbalanced datasets. Additionally, a web application was implemented using Flask to process consumption data and provide real-time predictions, optimizing the management of nontechnical losses. The results confirm that, with real consumption data, the algorithms achieve high accuracy in fraud detection, even in scenarios with severe class imbalance, validating the robustness of the system. Beyond its application in the energy sector, this solution is adaptable to other contexts requiring anomaly detection in transactional data. This proposal contributes to reducing economic losses, improving operational efficiency, and laying the groundwork for future research in intelligent monitoring systems., This work was partially supported by the Spanish Government under this research project funded by MCIN/AEI/10.13039/501100011033: DISCOVERY PID2023148716OB-C32. Also, by the project MultiMO TSI-1001232024-0060; and by the predoctoral scholarship associated with the "Generación de Conocimiento" Projects, Call 2022, PRE2021-099830. Also, by the Generalitat de Catalunya AGAUR grant "2021 SGR 01413"., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

On the adaptation of trajectory generative models to enforce differential privacy through gradient perturbation

  • Rubio Jornet, Víctor|||0009-0003-5977-2419
  • Parra Arnau, Javier|||0000-0002-1772-1088
  • Forné Muñoz, Jorge|||0000-0002-8401-3292
Synthetic trajectories have become a crucial tool for enabling data sharing in domains such as telecommunications and urban planning, yet most generators provide neither formal privacy guarantees nor empirical validation of the protection they claim. This paper investigates whether state-of-the-art deep learning trajectory generators can be adapted to satisfy differential privacy (DP) while retaining usable utility, and makes three contributions. First, we analyze 16 state-of-the-art generators and determine which can be trained with Differentially Private Stochastic Gradient Descent (DP-SGD), showing that common architectural choices — such as index-based coordinate look-ups at inference — silently void the guarantee. Second, we present Sec-LSTM-TrajGAN, a DP adaptation of LSTM-TrajGAN, a well-known trajectory generator chosen after the analysis of the existing models. Third, we introduce RouteMIA, to our knowledge the first closed-box membership inference attack tailored to synthetic trajectories, which adapts the distanceto-closest-record attack to sequential spatio-temporal data and requires access only to the released dataset. Evaluating on GeoLife (17, 000 trajectories) and Porto (100, 000 trajectories) datasets for several scenarios of increasing privacy. RouteMIA attains an AUC of 0.59 and 0.55 against the non-private models, which drops to ˜ 0.50 — chance level — for every private configuration. Notably, gradient clipping alone, with no noise added, already accounts for this entire empirical gain, at a utility cost of a distance JSD rising from 0.12 to 0.20 on Porto and essentially none on GeoLife, whereas strict budgets (e = 3) degrade GeoLife markedly (distance JSD up to 0.35). Per-example clipping, not the noise, therefore delivers the measurable protection against this attack — a cheap empirical complement to, but not a replacement for, formal guarantees., This work was supported in part by Spanish Government, funded by Ministerio de Ciencia e Innovación (MCIN)/Agencia Estatal de Investigación (AEI)/10.13039/501100011033 through the research project ‘‘DIstributed Smart COmmunications with Verifiable EneRgy-optimal Yields (DISCOVERY)’’ under Grant PID2023-148716OB-C32, through the research project ‘‘Privacy-Preserving Techniques for Mobility Data (PRIMOD)’’ under Grant CNS2025-166870, through the research project ‘‘Anonymization Technology for Artificial Intelligence (AI)-Based Analytics of Mobility Data (MOBILYTICS)’’ under Grant TED2021-129782B-I00, and through the research project ‘‘Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)’’ under Grant PID2020-113795RB-C31; in part by the project MOBILYTICS funded by European Union—NextGenerationEU/Recovery, Transformation and Resilience Plan (PRTR); and in part by the Generalitat de Catalunya through AGAUR under Grant 2021 SGR 01413. The work of Víctor Rubio-Jornet was supported by the Joan Oró Agència de Gestió d’Ajuts Universitaris i de Recerca (AGAUR)-Formació d’Investigadors (FI) Predoctoral Program, funded by the Secretariat of Universities and Research of the Department of Research and Universities of the Generalitat of Catalonia and European Social Fund Plus under Grant 2024 FI-1 00762. The work of Javier Parra-Arnau was supported by Ramón y Cajal Grant through Spanish Ministry of Science and Innovation and European Union—‘‘NextGenerationEU’’/PRTR under Grant RYC2021-034256-I., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Design of a SUMO-based simulator for optimal location of emergency vehicles in the Emergency Medical Systems

  • Mínguez, Antoni
  • Tozan, Volkan
  • El-Assaad, Monica
  • Sola Muñoz, Silvia
  • Azeli, Youcef
  • Cruz Llopis, Luis Javier de la|||0000-0003-4171-8310
  • Jiménez-Fàbrega, Xavier
  • Galup, Carles
  • Dalmau, Raimon
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
Te response time between an accident happening and an ambulance reaching the incident place is a very delicate andcrucial time. Te time elapsed between those two events could be the determining factor of whether the injured person getsto live or die. Terefore, the response time of the ambulance assisting in the accident should be as small as possible. Responsetime can directly afect a person’s recovery time and can make the diference between the patient sufering temporary orprolonged disability, or even death. Locating the best position for ambulance bases is a challenge for emergency medicalsystems (EMSs). Tis will improve the EMS service in assigning the ambulance that arrives sooner to a given incident, thusreducing the risks associated with long response times. In this paper, we design a SUMO-based simulation frameworknamed SEMSIM (Sistema d’Emerg`encies M`ediques SIMulator) that mimics the management of ambulances done by theEMS. Te goal is to develop a tool to easily explore alternatives of stationing the available ambulances at the available bases,following diferent confguration settings (i.e., distributing the ambulances in diferent ways in the current bases or in-cluding new base locations), and analyze the average response time for each of those alternative confgurations incomparison to the current confguration. Te designed tool will help the EMS to explore dynamic confgurations that areoptimal throughout time, e.g. for diferent seasons, type of day, or time of the day. Te SEMSIM simulation tool accuratelyreplicates the ambulance assignments and response times of the EMS, validating its efectiveness. Tis allows SEMSIM to testnew ambulance distribution confgurations, including the addition of new bases, which would be costly or impractical toexplore in real life. SEMSIM provides a valuable resource for EMS to optimize their operations using real incident data andmaps from OpenStreetMap. Te aim in this paper is to design a methodology based on dynamic SUMO simulations toanalyze the optimal dynamic confguration of the feet of ambulances that ofers the smallest average response time and thushelps to improve the EMS service. As a case study, we explored various alternative confgurations reducing response time bymore than 2 min., This work was partially supported by the Spanish Government under research projects: (i) “DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)” PID2023-148716OB-C32, funded by MCIN/AEI/10.13039/501100011033; (ii) “Anonymization technology for AI-based analytics of mobility data (MOBILYTICS)” TED2021-129782B-I00, funded by MCIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/ PRTR; (iii) “Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)” PID2020-113795RB-C31, funded by MCIN/AEI/10.13039/501100011033; and (iv) the Generalitat de Catalunya under AGAUR grant “2021 SGR 01413”., Peer Reviewed, Objectius de Desenvolupament Sostenible::3 - Salut i Benestar
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Federated learning-based electric vehicle energy consumption prediction and charging station recommendation

  • Al-zuhairi, Yaqoob
  • Ali, Aya Maher
  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
Electric vehicles (EVs) offer significant potential for reducing emissions, yet their expansion is constrained by long charging times, limited charging infrastructure, and inefficient charging station (CS) utilization. This study proposes an intelligent platform that explicitly supports drivers in multiple aspects: predicting EV energy consumption (EVEC), estimating the remaining energy at the destination, and determining the remaining driving range, thereby assisting drivers in deciding whether to continue their trip or stop for recharging. The system also recommends the most appropriate CS by considering driver preferences. Using SUMO simulations with OpenStreetMap data to prepare realistic traffic scenarios, the platform combines ensemble machine learning (ML) models with federated learning (FL) to optimize EV charging decisions. Through the integration of Bi-LSTM + XGBoost for EVEC prediction and FFNN + XGBoost for the optimal CS selection, the proposed framework significantly outperforms conventional methods by minimizing total travel time and resulting in mean absolute error (MAE) values ranging from 2.3 to 4.5 minutes under varying traffic conditions., This work was partially supported by the Spanish Government under research projects: (i) DISCOVERY PID2023-148716OB-C32, funded by MCIN/AEI/10.13039/501100011033; (ii) the Generalitat de Catalunya under AGAUR grant ”2021 SGR 01413”; (iii) MultiMO project TSI-100123-2024-60; and (iv) Predoctoral Scholarship under Grant PRE2021-099., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Simulation under stress: a comparative benchmarking of large-scale traffic simulators

  • Guin, Agneev
  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Kannan, Prashanth
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
Modern urban mobility systems are increasingly dependent on precise and scalable simulation platforms to facilitate the design, assessment, and optimization of intelligent transportation systems (ITS). This paper sets forth a thorough benchmarking investigation of three extensively utilized traffic simulation platforms: SUMO (Simulation of Urban MObility), CityFlow, and MATSim, with an emphasis on their computational efficacy within large-scale synthetic traffic scenarios. We conduct an evaluation of these platforms across vehicle volumes ranging from 10 to 10 million vehicles, assessing total runtime, CPU and GPU utilization, and memory consumption on high-performance computing infrastructures. Our results reveal notable architectural trade-offs: SUMO exhibits predictable linear scaling but becomes constrained by CPU limitations at elevated vehicle counts, CityFlow encounters memory limitations beyond 10,000 vehicles, whereas MATSim necessitates meticulous JVM (Java Virtual Machine) optimization to effectively manage large-scale scenarios. This investigation provides essential guidance for researchers and practitioners in the selection of suitable simulation tools for urban-scale traffic modeling and identifies critical computational challenges associated with the scaling of simulations for smart city initiatives., This work was supported by the grant PID2023-148716OB-C31 funded by MCIU/AEI/10.13039/501100011033 (DISCOVERY project). Also, by the project ”MultiMO: Datos MultiSectoriales para la Movilidad Obligada” TSI100123-2024-60 (European Union, NextGenerationEU); Also, by predoctoral scholarship for the training of research personnel associated with the ”Generaci´on de Conocimiento” Project PRE2021-099830; Finally, by the Generalitat de Catalunya under AGAUR grant ”2021 SGR 01413”., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

Comparing optimal and adaptive EV charging in smart cities: MILP vs. reinforcement learning

  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Barbecho Bautista, Pablo
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
  • Cuomo, Francesca
The coordinated scheduling of electric vehicle (EV) charging is a critical challenge for smart cities, particularly in high-density infrastructure such as Mobility Hubs (MHs). This paper evaluates and compares two prominent approaches to the EV Charging Scheduling Problem (CSP): Mixed-Integer Linear Programming (MILP) and Reinforcement Learning (RL). We formulate a shared problem framework and apply both strategies under two structured scenarios: a small-scale deterministic benchmark and a medium-scale, realistic deployment with higher heterogeneity. Results show that MILP achieves optimal cost and state of charge SoC compliance in tractable cases but struggles with scalability. RL, based on Proximal Policy Optimization (PPO), achieves near-optimal performance while scaling to 100 EVs with minimal computation time. Despite occasional SoC deviations, the RL agent exhibits robust and adaptive behavior under dynamic conditions. This study offers actionable insights for selecting and deploying EV scheduling strategies in real-world urban environments., This work was partially supported by the Spanish Government under this research project funded by MCIN/AEI/10.13039/501100011033: DISCOVERY PID2023-148716OB-C32. Also by the project MultiMO TSI100123-2024-0060; and by the predoctoral scholarship associated with the ”Generaci´on de Conocimiento” Projects, Call 2022, PRE2021-099830. Also, by the Generalitat de Catalunya AGAUR grant ”2021 SGR 01413”., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Contribución de congreso (ConferenceOutput)

RUTGe: realistic urban traffic generator for urban environments using deep reinforcement learning and SUMO simulator

  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Barbecho Bautista, Pablo
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
We are witnessing a profound shift in societal and political attitudes, driven by the visible consequences of climate change in urban environments. Urban planners, public transport providers, and traffic managers are urgently reimagining cities to promote sustainable mobility and expand green spaces for pedestrians, bicycles, and scooters. To design more sustainable cities, urban planners require realistic simulation tools to optimize mobility, identify location for car chargers, convert streets to pedestrian zones, and evaluate the impact of alternative configurations. However, realistic traffic profiles are essential to produce meaningful simulation results. Addressing this need, we propose a traffic generator based on deep reinforcement learning integrated with the SUMO simulator. This tool learns to generate an instantaneous number of vehicles throughout the day, aligning closely with the target profiles observed at the traffic monitoring stations. Our approach generates accurate 24-hour traffic patterns for any city using minimal statistical data, achieving higher accuracy compared to existing alternatives. In particular, our proposal demonstrates a highly accurate 24-hour traffic adjustment, with the generated traffic deviating only by about 5% from the real target traffic. This performance significantly exceeds that of current SUMO tools like RouteSampler, which struggle to accurately follow the total daily traffic curve, especially during peak hours when severe traffic congestion occurs., This work was partially supported by the Spanish Government under these research projects funded by MCIN/AEI/10.13039/501100011033: DISCOVERY PID2023-148716OB-C32; MOBILYTICS TED2021-129782B-I00 (also funded by the European Union NextGenerationEU/PRTR); COMPROMISE PID2020-113795RB-C31; predoctoral scholarship associated with the ”Generaci´ on de Conocimiento” Projects, Call 2022, PRE2021099830. Also, by the Generalitat de Catalunya AGAURgrant ”2021 SGR 01413”., Peer Reviewed, 11 - Ciutats i Comunitats Sostenibles, 11.2 - Per a 2030, proporcionar accés a sistemes de transport segurs, assequibles, accessibles i sostenibles per a totes les persones, i millorar la seguretat viària, en particular mitjançant l’ampliació del transport públic, amb especial atenció a les necessitats de les persones en situació vulnerable, dones, nenes, nens, persones amb discapacitat i persones grans
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

FedCAMO: Federated learning carbon-aware multi-objective client selection

  • Junjun, Lu
  • Postigo Boix, Marcos|||0000-0002-4182-7376
  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Cruz Llopis, Luis Javier de la|||0000-0003-4171-8310
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
This work was supported by the project ‘‘[DISCOVERY]: Distributed Smart Communications with Verifiable EneRgy-optimal Yields’’ PID2023-148716OB-C32 (Agencia Estatal de Investigación, Spain, Ministerio de Ciencia e Innovación); also by the project ‘‘MultiMO: Datos MultiSectoriales para la Movilidad Obligada’’ TSI-100123-2024-60 (Ministerio de transformación digital y de la función pública, NextGenerationEU); also, by predoctoral scholarship for the training of research personnel associated with the ‘‘Generación de Conocimiento’’ Project PRE2021-099830., Federated Learning (FL) enables collaborative model training on geographically dispersed devices, offering significant privacy and efficiency benefits. However, large-scale FL deployment faces a critical sustainability challenge due to potentially higher energy consumption and carbon emissions. Current FL schedulers fail to effectively balance three conflicting objectives: (i) model performance, (ii) data diversity, and (iii) carbon emissions. Solely minimizing carbon degrades convergence, while ignoring it results in an unsustainable environmental footprint. The core challenge is therefore to maximize system utility under a strict carbon budget while preserving data diversity across clients with heterogeneous hardware and geographical carbon intensity. To address this issue, we propose FedCAMO (Federated Learning Carbon-Aware Multi-Objective Client Selection), a framework modeled as a Combinatorial Multi-Armed Bandit (CMAB) problem subject to per-round constraints. FedCAMO enables the server to learn client characteristics in real-time—accounting for heterogeneous hardware and uneven geographical Carbon Intensity (CI)—to select cohorts that maximize system utility under strict per-round carbon constraints. We integrate a holistic, system-wide energy model that, unlike existing work, accounts for the energy consumption of all participants—including the central server—during both active and idle states. Algorithmically, FedCAMO employs an online dual-variable algorithm to learn a dynamic “carbon price”. This allows the system to adaptively regulate selection pressure, ensuring that training strictly respects the carbon budget while optimizing the trade-off between accuracy and diversity. We implemented FedCAMO within the Flower framework, demonstrating its practical portability. Extensive simulations demonstrate that FedCAMO strictly respects per-round carbon budgets while achieving a significantly superior Pareto-optimal frontier compared to carbon-blind and simplified carbon-aware baselines. Specifically, FedCAMO improves final accuracy by up to 1.9 percentage points over the strongest competing baseline. In highly carbon-intensive environments, it reduces total COe by approximately 50% (and up to 70% on specific workloads) relative to FedAvg while simultaneously improving model accuracy. These results indicate that carbon-aware control need not sacrifice learning quality when utility, diversity, and holistic carbon cost are coordinated explicitly., Peer Reviewed, 7 - Energia Assequible i No Contaminant, 7.b - Per a 2030, ampliar la infraestructura i millorar la tecnologia per tal d’oferir serveis d’energia moderns i sostenibles per a tots els països en desenvolupament, en particular els països menys avançats, els petits estats insulars en desenvolupament i els països en desenvolupament sense litoral, d’acord amb els programes de suport respectius
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Deployment-aware multi-objective optimization of machine learning models for predictive congestion control in multi-hop vehicular networks

  • Astudillo León, Juan Pablo|||0000-0002-7291-3584
  • Lemus Cárdenas, Leticia
  • Cruz Llopis, Luis Javier de la|||0000-0003-4171-8310
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
Multi-hop vehicular wireless networks are highly susceptible to channel congestion due to dynamic topologies, heterogeneous traffic densities, and contention-based medium access. This paper proposes a deployment-aware predictive congestion control framework for vehicular communications, where packet transmission decisions are guided by supervised machine learning models optimized under predictive and computational constraints. Its novelty lies in coupling packet-level congestion prediction with deployment-aware model selection, allowing transmission-control models to be selected according to predictive quality, measured native-inference latency, and serialized model footprint. A packet-level dataset was constructed from realistic vehicle trajectories generated with SUMO using road networks extracted from OpenStreetMap for three topologically diverse cities: Liverpool, Rio de Janeiro, and Houston. These trajectories support vehicular network evaluation under heterogeneous urban conditions, enabling the assessment of predictive transmission decisions across different road topologies. Several supervised learning models were considered, including linear classifiers, neural networks, and ensemble machine learning architectures such as Random Forest, Weighted Soft Voting, and Stacking. Hyperparameter tuning was formulated as a multi-objective optimization problem and solved using the NSGA-II evolutionary algorithm, jointly maximizing predictive performance while minimizing inference latency and serialized model size. Pareto-efficient models were exported to ONNX format and executed through a native inference pipeline to evaluate their suitability for computationally constrained vehicular communication environments. Results across heterogeneous urban scenarios show that the proposed framework improves vehicular communication performance in terms of packet delivery reliability and communication delay, while preserving competitive machine learning performance under deployment-oriented constraints., This work was supported in part by Spanish Government funded by MCIN/AEI/10.13039/501100011033 under Research Project ‘‘DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)’’ under Grant PID2023-148716OB-C32. The work of Juan Pablo Astudillo León was supported by Yachay Tech University for the support given to the ‘‘VICOM-AI: Vehicular Intelligent Communications Optimization Using Artificial Intelligence Project’’ under Grant MATH25-05., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

DNS query forgery: a client-side defense against mobile app traffic profiling

  • Jiménez Berenguel, Andrea
  • Gil Espinasa, César
  • García Rubio, Carlos
  • Forné Muñoz, Jorge|||0000-0002-8401-3292
  • Campo Vazquez, Celeste
Mobile applications generate DNS queries that expose user behavioral patterns to network observers, creating privacy vulnerabilities even when communications are encrypted. Network eavesdroppers and DNS resolvers can analyze domain name sequences to profile users based on their app usage patterns. This paper proposes a client-side defense mechanism based on DNS query forgery to obfuscate user DNS-based profiles. Our method applies a query forgery technique that consists of injecting false DNS queries into genuine traffic streams. We mathematically model user profiles as probability distributions over interest categories and analyze the optimal proportion of false queries needed to achieve desired privacy levels. We evaluate three query forgery strategies: Uniform, TrackMeNot-based, and Optimized, finding that the Optimized strategy using KL divergence is the most effective. To validate our approach, we develop a novel methodology for generating synthetic user traces, creating a dataset of 1,000 users by mapping real app traffic data onto individual user profiles. Our analysis reveals that 50% privacy improvement is achievable with less than 20% traffic overhead, while 100% privacy protection requires approximately 40-60% additional traffic. We further propose a modular system architecture for practical implementation on mobile devices. This work offers a client-side privacy solution that operates without third-party trust requirements, empowering users to defend against traffic analysis without compromising application functionality., This work was supported in part by COMPROMISE Project through Ministerio de Ciencia, Innovación y Universidades (MICIU)/Agencia Estatal de Investigación (AEI)/10.13039/501100011033 under Grant PID2020-113795RB-C32 and Grant PID2020-113795RB-C31; in part by QURSA Project through MICIU/AEI/10.13039/501100011033 and European Union NextGenerationEU/Recovery, Transformation and Resilience Plan (PRTR) under Grant TED2021-130369B-C32; in part by MOBILYTICS Project through MICIU/AEI/10.13039/501100011033 and European Union NextGenerationEU/PRTR under Grant TED2021-129782B-I00; in part by SISCOM Project funded by the Generalitat de Catalunya through the Agéncia de Gestió d’Ajuts Universitaris i de Recerca (AGAUR) under Grant 2021 SGR 01413; in part by DISCOVERY Project through MICIU/AEI/10.13039/501100011033 under Grant PID2023-148716OB-C33 and Grant PID2023-148716OB-C32; and in part by Fondo Europeo de Desarrollo Regional (FEDER), Union Europea (UE), and the I-Shaper Strategic Project, a collaboration agreement between the Instituto Nacional de Ciberseguridad (INCIBE) and the Universidad Carlos III de Madrid, an initiative framework of PRTR through European Union (Next Generation) under Grant C114/23., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

MobiCugat: city-scale traffic assessment using low-emission zone camera data

  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Rubio Jornet, Víctor|||0009-0003-5977-2419
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
  • Montal, Joaquim
  • Vives, Marta
  • Muratet, Albert
While Low Emission Zone (LEZ) enforcement cameras provide a constant stream of traffic data, such resources remain significantly underexploited for urban mobility planning, as their current application is restricted to enforcing vehicle access regulations and issuing fines. This paper presents MobiCugat, a framework demonstrating that Automatic Number Plate Recognition (ANPR) camera data from a municipal LEZ network can serve as the calibration backbone for high-fidelity, city-scale traffic simulations for a policy-testing Digital Twin. The case study is Sant Cugat del Vallès (Barcelona), where the local council sought to evaluate new scenarios for the area using an evidence-based, data-driven approach. Vehicle detection records from 102 LEZ ANPR cameras were processed into 15-min traffic intensity time series through a General Data Protection Regulation (GDPR)-compliant pipeline. The Realistic Urban Traffic Generator (RUTGe), a Deep Reinforcement Learning-based tool, was used to generate SUMO-compatible traffic demand whose simulated detector counts reproduce the observed camera-based intensities. The resulting simulations reproduced the observed detector-level traffic intensities with MARE% values between 2.29% and 2.90% across representative morning peak, midday off-peak, and evening peak traffic conditions. Additionally, camera analysis of over 470,000 vehicle records revealed that resident traffic (37.4%) dominates over through-traffic (3.8%), significantly refining prior survey-based estimates. Our high-fidelity simulation tool based on SUMO, features realistic traffic patterns calibrated through AI-driven techniques, enabling the evaluation of diverse ’what-if’ scenarios—such as road closures, pedestrianization, changes in traffic direction, or relocation of bus stops. By quantifying the impact of these interventions, our tool facilitates informed decision-making prior to physical implementation. The proposed pipeline is cost-effective, privacy-preserving, and directly replicable for any municipality operating an LEZ camera network, offering a scalable template for evidence-based urban mobility planning, aligned with the European Strategy for Data and the EU Green Deal goals for sustainable mobility., This work was supported by the project “MobiCugat: Estudi de mobilitat urbana per l’Ajuntament de Sant Cugat” (C-13469); by the project “DISCOVERY: DIstributed Smart Communications with Verifiable EneRgy-optimal Yields” PID2023-148716OB-C32 (Agencia Estatal de Investigación, Ministerio de Ciencia e Innovación); by the project “MultiMO: Datos MultiSectoriales para la Movilidad Obligada” TSI-100123-2024-60 (Ministerio de transformación digital y de la función pública, NextGenerationEU); also by doctoral scholarship PRE2021-099830 funded by the Ministerio de Ciencia, Innovación y Universidades; and by doctoral scholarship Joan Oró “2024 FI-1 00762” funded by the Agència de Gestió d’Ajuts Universitaris i de Recerca (AGAUR)., Peer Reviewed, 11 - Ciutats i Comunitats Sostenibles, 11.2 - Per a 2030, proporcionar accés a sistemes de transport segurs, assequibles, accessibles i sostenibles per a totes les persones, i millorar la seguretat viària, en particular mitjançant l’ampliació del transport públic, amb especial atenció a les necessitats de les persones en situació vulnerable, dones, nenes, nens, persones amb discapacitat i persones grans
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Uncoordinated syntactic privacy: a new composable metric for multiple, independent data publishing

  • Tobar Nicolau, Adrián|||0000-0003-0198-2475
  • Parra Arnau, Javier|||0000-0002-1772-1088
  • Forné Muñoz, Jorge|||0000-0002-8401-3292
  • Torra, Vicenç
A privacy model is a privacy condition, dependent on a parameter, that guarantees an upper bound on the risk of reidentification disclosure and maybe also on the risk of attribute disclosure by an adversary. A privacy model is composable if the privacy guarantees of the model are preserved, possibly to a limited extent, after repeated independent application of the privacy model. From the opposite perspective, a privacy model is not composable if multiple independent data releases, each of them satisfying the requirements of the privacy model, may result in a privacy breach. Current privacy models are broadly classified into syntactic ones (such as k-anonymity and l-diversity) and semantic ones, which essentially refer to e -differential privacy (e-DP) and variations thereof. While e-DP and its variants offer strong composability properties, syntactic notions are not composable unless data releases are conducted by a single, centralized data holder that uses specialized notions such as m-invariance and t -safety. In this work, we propose m-uncoordinated-syntactic-privacy (m-USP), the first syntactic notion with composability properties for the independent publication of nondisjoint data, in other words, without a centralized data holder. Theoretical results are formally proven, and experimental results demonstrate that the risk to individuals does not increase significantly, in contrast to non-composable methods, that are susceptible to attribute disclosure. In most cases, the utility degradation caused by the extra protection is less than 5% and decreases as the value of m increases., Javier Parra-Arnau is the recipient of a “Ram´ on y Cajal” fellowship (ref. RYC2021-034256-I) funded by the Spanish Ministry of Science and Innovation and the European Union– “NextGenerationEU”/PRTR (Plan de Recuperaci´ on, Transformaci´ on y Resiliencia). This work was also supported by the Spanish Government under the following projects: ”DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)” PID2023-148716OB-C32, funded by MCIN/AEI/10.13039/501100011033; “Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)” PID2020-113795RBC31, funded by MICIU/AEI/10.13039/501100011033; and “Anonymization technology for AI-based analytics of mobility data (MOBILYTICS)” (TED2021-129782B-I00), funded by MICIU/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR. This work was also partially funded by the Generalitat de Catalunya, under AGAUR grant “2021 SGR 01413”; was partially supported by the Wallenberg AI, Autonomous Systems and Software Program (WASP) funded by the Knut and Alice Wallenberg Foundation; and received support by the project ”Privacy for complex data” (VR 2022-04645).
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

Efficient charging station selection for minimizing total travel time of electric vehicles

  • Al-zuhairi, Yaqoob
  • Kannan, Prashanth
  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Cruz Llopis, Luis Javier de la|||0000-0003-4171-8310
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
Electric vehicles (EVs) have gained significant attention in recent decades for their environmental benefits. However, their widespread adoption poses challenges due to limited charging infrastructure and long charging times, often resulting in underutilized charging stations (CSs) and unnecessary queues that complicate travel planning. Therefore, selecting the appropriate CS is essential for minimizing the total travel time of EVs, as it depends on both driving time and the required charging duration. This selection process requires estimating the energy required to reach each candidate CS and then continue to the destination, while also checking if the EV’s battery level is sufficient for a direct trip. To address this gap, we propose an integrated platform that leverages two ensemble machine learning models: Bi-LSTM + XGBoost to predict energy consumption, and FFNN + XGBoost for identifying the most suitable CS by considering required energy, waiting time at CS, charging speed, and driving time based on varying traffic conditions. This integration forms the core novelty of our system to optimize CS selection to minimize the total trip duration. This approach was validated with SUMO simulations and OpenStreetMap data, demonstrating a mean absolute error (MAE) ranging from 2.29 to 4.5 min, depending on traffic conditions, outperforming conventional approaches that rely on SUMO functions and mathematical calculations, which typically yielded MAEs between 5.1 and 10 min. These findings highlight the proposed system’s effectiveness in reducing total travel time, improving charging infrastructure utilization, and enhancing the overall experience for EV drivers., This work was partially supported by the Spanish Government under research projects: (i) “DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)” PID2023-148716OB-C32, funded by MCIN/AEI/10.13039/501100011033; (ii) “Anonymization technology for AI-based analytics of mobility data (MOBILYTICS)” TED2021-129782B-I00, funded by MCIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/ PRTR; (iii) “Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)”PID2020-113795RB-C31, fundedbyMCIN/AEI/10.13039/501100011033; (iv)the Generalitat de Catalunya under AGAUR grant “2021 SGR 01413”; and (v) “MultiMO: Datos MultiSectoriales para la Movilidad Obligada” TSI-100123-2024-60 (European Union, NextGenerationEU)., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publicación Artículo científico (JournalArticle)

FedITS-tool: A lockstep co-simulation framework for vehicular federated learning with per-round energy and carbon accounting

  • Junjun, Lu
  • Postigo Boix, Marcos|||0000-0002-4182-7376
  • Bazán Guillén, Alberto|||0000-0001-8634-6907
  • Cruz Llopis, Luis Javier de la|||0000-0003-4171-8310
  • Aguilar Igartua, Mónica|||0000-0002-6518-888X
Vehicular federated learning (VFL) under realistic mobility should be evaluated not only by model accuracy, but also by the energy and carbon cost of distributed training. Mobility, intermittent connectivity, and deadline misses can cause clients to consume energy for updates that are later dropped before aggregation, producing wasted carbon emissions. This paper presents FedITS-Tool, a modular lockstep co-simulation framework that couples SUMO-based vehicle mobility, a lightweight Veins/OMNeT++ connectivity-time service, and Flower-based FL training in Docker containers. A central orchestrator synchronizes simulation time with FL rounds, maps available clients to simulated vehicles, enforces deadline-aware commit/drop semantics, and exports per-round traces of failure causes, communication delays, energy use, and carbon emissions. Experiments show that FedITS-Tool quantifies up to 40% dropped-carbon waste in coverage-stressed scenarios while preserving traceability in 500-vehicle simulations. The tool supports reproducible studies of RSU coverage planning, deadline analysis, participation dynamics, and carbon-aware VFL design., This work was supported by the project “DISCOVERY: Distributed Smart Communications with Verifiable Energy-optimal Yields” PID2023-148716OB-C32 (Agencia Estatal de Investigación, Ministerio de Ciencia e Innovación); also by the project “MultiMO: Datos MultiSectoriales para la Movilidad Obligada” TSI-100123-2024-60 (Ministerio de transformación Digital y de la función pública, NextGenerationEU); also, by a predoctoral scholarship for the training of research personnel associated with the “Generación de Conocimiento” Project PRE2021-099830., Peer Reviewed
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

DNS query forgery: a client-side defense against mobile app traffic profiling

  • Jiménez Berenguel, Andrea
  • Gil Espinasa, César
  • García Rubio, Carlos
  • Forné Muñoz, Jorge
  • Campo Vazquez, Celeste
Mobile applications generate DNS queries that expose user behavioral patterns to network observers, creating privacy vulnerabilities even when communications are encrypted. Network eavesdroppers and DNS resolvers can analyze domain name sequences to profile users based on their app usage patterns. This paper proposes a client-side defense mechanism based on DNS query forgery to obfuscate user DNS-based profiles. Our method applies a query forgery technique that consists of injecting false DNS queries into genuine traffic streams. We mathematically model user profiles as probability distributions over interest categories and analyze the optimal proportion of false queries needed to achieve desired privacy levels. We evaluate three query forgery strategies: Uniform, TrackMeNot-based, and Optimized, finding that the Optimized strategy using KL divergence is the most effective. To validate our approach, we develop a novel methodology for generating synthetic user traces, creating a dataset of 1,000 users by mapping real app traffic data onto individual user profiles. Our analysis reveals that 50% privacy improvement is achievable with less than 20% traffic overhead, while 100% privacy protection requires approximately 40-60% additional traffic. We further propose a modular system architecture for practical implementation on mobile devices. This work offers a client-side privacy solution that operates without third-party trust requirements, empowering users to defend against traffic analysis without compromising application functionality., This work was supported in part by COMPROMISE Project through Ministerio de Ciencia, Innovación y Universidades (MICIU)/Agencia Estatal de Investigación (AEI)/10.13039/501100011033 under Grant PID2020-113795RB-C32 and Grant PID2020-113795RB-C31; in part by QURSA Project through MICIU/AEI/10.13039/501100011033 and European Union NextGenerationEU/Recovery, Transformation and Resilience Plan (PRTR) under Grant TED2021-130369B-C32; in part by MOBILYTICS Project through MICIU/AEI/10.13039/501100011033 and European Union NextGenerationEU/PRTR under Grant TED2021-129782B-I00; in part by SISCOM Project funded by the Generalitat de Catalunya through the Agéncia de Gestió d’Ajuts Universitaris i de Recerca (AGAUR) under Grant 2021 SGR 01413; in part by DISCOVERY Project through MICIU/AEI/10.13039/501100011033 under Grant PID2023-148716OB-C33 and Grant PID2023-148716OB-C32; and in part by Fondo Europeo de Desarrollo Regional (FEDER), Union Europea (UE), and the I-Shaper Strategic Project, a collaboration agreement between the Instituto Nacional de Ciberseguridad (INCIBE) and the Universidad Carlos III de Madrid, an initiative framework of PRTR through European Union (Next Generation) under Grant C114/23., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Efficient charging station selection for minimizing total travel time of electric vehicles

  • Al-zuhairi, Yaqoob
  • Kannan, Prashanth
  • Bazán Guillén, Alberto
  • Cruz Llopis, Luis Javier de la
  • Aguilar Igartua, Mónica
Electric vehicles (EVs) have gained significant attention in recent decades for their environmental benefits. However, their widespread adoption poses challenges due to limited charging infrastructure and long charging times, often resulting in underutilized charging stations (CSs) and unnecessary queues that complicate travel planning. Therefore, selecting the appropriate CS is essential for minimizing the total travel time of EVs, as it depends on both driving time and the required charging duration. This selection process requires estimating the energy required to reach each candidate CS and then continue to the destination, while also checking if the EV’s battery level is sufficient for a direct trip. To address this gap, we propose an integrated platform that leverages two ensemble machine learning models: Bi-LSTM + XGBoost to predict energy consumption, and FFNN + XGBoost for identifying the most suitable CS by considering required energy, waiting time at CS, charging speed, and driving time based on varying traffic conditions. This integration forms the core novelty of our system to optimize CS selection to minimize the total trip duration. This approach was validated with SUMO simulations and OpenStreetMap data, demonstrating a mean absolute error (MAE) ranging from 2.29 to 4.5 min, depending on traffic conditions, outperforming conventional approaches that rely on SUMO functions and mathematical calculations, which typically yielded MAEs between 5.1 and 10 min. These findings highlight the proposed system’s effectiveness in reducing total travel time, improving charging infrastructure utilization, and enhancing the overall experience for EV drivers., This work was partially supported by the Spanish Government under research projects: (i) “DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)” PID2023-148716OB-C32, funded by MCIN/AEI/10.13039/501100011033; (ii) “Anonymization technology for AI-based analytics of mobility data (MOBILYTICS)” TED2021-129782B-I00, funded by MCIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/ PRTR; (iii) “Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)”PID2020-113795RB-C31, fundedbyMCIN/AEI/10.13039/501100011033; (iv)the Generalitat de Catalunya under AGAUR grant “2021 SGR 01413”; and (v) “MultiMO: Datos MultiSectoriales para la Movilidad Obligada” TSI-100123-2024-60 (European Union, NextGenerationEU)., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Federated learning-based electric vehicle energy consumption prediction and charging station recommendation

  • Al-zuhairi, Yaqoob
  • Ali, Aya Maher
  • Bazán Guillén, Alberto
  • Aguilar Igartua, Mónica
Electric vehicles (EVs) offer significant potential for reducing emissions, yet their expansion is constrained by long charging times, limited charging infrastructure, and inefficient charging station (CS) utilization. This study proposes an intelligent platform that explicitly supports drivers in multiple aspects: predicting EV energy consumption (EVEC), estimating the remaining energy at the destination, and determining the remaining driving range, thereby assisting drivers in deciding whether to continue their trip or stop for recharging. The system also recommends the most appropriate CS by considering driver preferences. Using SUMO simulations with OpenStreetMap data to prepare realistic traffic scenarios, the platform combines ensemble machine learning (ML) models with federated learning (FL) to optimize EV charging decisions. Through the integration of Bi-LSTM + XGBoost for EVEC prediction and FFNN + XGBoost for the optimal CS selection, the proposed framework significantly outperforms conventional methods by minimizing total travel time and resulting in mean absolute error (MAE) values ranging from 2.3 to 4.5 minutes under varying traffic conditions., This work was partially supported by the Spanish Government under research projects: (i) DISCOVERY PID2023-148716OB-C32, funded by MCIN/AEI/10.13039/501100011033; (ii) the Generalitat de Catalunya under AGAUR grant ”2021 SGR 01413”; (iii) MultiMO project TSI-100123-2024-60; and (iv) Predoctoral Scholarship under Grant PRE2021-099., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Artificial intelligence methods for anomaly detection in energy consumption

  • Pérez Méndez, Ana Laura
  • Monteagudo Gordillo, René
  • Bazan Prieto, Carlos Alberto
  • Bazán Guillén, Alberto
  • Bello Pérez, Rafael E.
  • Aguilar Igartua, Mónica
Non-technical losses represent a significant challenge for energy distribution companies due to their economic impact. These losses typically arise from irregularities at supply points and fraudulent customer behavior. In Cuba, electricity meter readings are performed manually, and consumption data is processed using spreadsheets that combine weighted criteria to generate alerts for potential anomalies. This procedure is prone to vulnerabilities such as manual data entry errors, incorrect key assignments, and human mistakes made by field readers, which compromise the reliability of the analysis. In this paper we present an anomaly detection system for electricity consumption, developed using artificial intelligence techniques to identify irregularities based on monthly reports from residential users. Various machine learning methods were evaluated, with eXtreme Gradient Boosting (XGBoost) standing out for its effectiveness in handling imbalanced datasets. Additionally, a web application was implemented using Flask to process consumption data and provide real-time predictions, optimizing the management of nontechnical losses. The results confirm that, with real consumption data, the algorithms achieve high accuracy in fraud detection, even in scenarios with severe class imbalance, validating the robustness of the system. Beyond its application in the energy sector, this solution is adaptable to other contexts requiring anomaly detection in transactional data. This proposal contributes to reducing economic losses, improving operational efficiency, and laying the groundwork for future research in intelligent monitoring systems., This work was partially supported by the Spanish Government under this research project funded by MCIN/AEI/10.13039/501100011033: DISCOVERY PID2023148716OB-C32. Also, by the project MultiMO TSI-1001232024-0060; and by the predoctoral scholarship associated with the "Generación de Conocimiento" Projects, Call 2022, PRE2021-099830. Also, by the Generalitat de Catalunya AGAUR grant "2021 SGR 01413"., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Design of a SUMO-based simulator for optimal location of emergency vehicles in the Emergency Medical Systems

  • Mínguez, Antoni
  • Tozan, Volkan
  • El-Assaad, Monica
  • Sola Muñoz, Silvia
  • Azeli, Youcef
  • Cruz Llopis, Luis Javier de la
  • Jiménez-Fàbrega, Xavier
  • Galup, Carles
  • Dalmau, Raimon
  • Aguilar Igartua, Mónica
Te response time between an accident happening and an ambulance reaching the incident place is a very delicate andcrucial time. Te time elapsed between those two events could be the determining factor of whether the injured person getsto live or die. Terefore, the response time of the ambulance assisting in the accident should be as small as possible. Responsetime can directly afect a person’s recovery time and can make the diference between the patient sufering temporary orprolonged disability, or even death. Locating the best position for ambulance bases is a challenge for emergency medicalsystems (EMSs). Tis will improve the EMS service in assigning the ambulance that arrives sooner to a given incident, thusreducing the risks associated with long response times. In this paper, we design a SUMO-based simulation frameworknamed SEMSIM (Sistema d’Emerg`encies M`ediques SIMulator) that mimics the management of ambulances done by theEMS. Te goal is to develop a tool to easily explore alternatives of stationing the available ambulances at the available bases,following diferent confguration settings (i.e., distributing the ambulances in diferent ways in the current bases or in-cluding new base locations), and analyze the average response time for each of those alternative confgurations incomparison to the current confguration. Te designed tool will help the EMS to explore dynamic confgurations that areoptimal throughout time, e.g. for diferent seasons, type of day, or time of the day. Te SEMSIM simulation tool accuratelyreplicates the ambulance assignments and response times of the EMS, validating its efectiveness. Tis allows SEMSIM to testnew ambulance distribution confgurations, including the addition of new bases, which would be costly or impractical toexplore in real life. SEMSIM provides a valuable resource for EMS to optimize their operations using real incident data andmaps from OpenStreetMap. Te aim in this paper is to design a methodology based on dynamic SUMO simulations toanalyze the optimal dynamic confguration of the feet of ambulances that ofers the smallest average response time and thushelps to improve the EMS service. As a case study, we explored various alternative confgurations reducing response time bymore than 2 min., This work was partially supported by the Spanish Government under research projects: (i) “DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)” PID2023-148716OB-C32, funded by MCIN/AEI/10.13039/501100011033; (ii) “Anonymization technology for AI-based analytics of mobility data (MOBILYTICS)” TED2021-129782B-I00, funded by MCIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/ PRTR; (iii) “Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)” PID2020-113795RB-C31, funded by MCIN/AEI/10.13039/501100011033; and (iv) the Generalitat de Catalunya under AGAUR grant “2021 SGR 01413”., Peer Reviewed, Objectius de Desenvolupament Sostenible::3 - Salut i Benestar, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Simulation under stress: a comparative benchmarking of large-scale traffic simulators

  • Guin, Agneev
  • Bazán Guillén, Alberto
  • Kannan, Prashanth
  • Aguilar Igartua, Mónica
Modern urban mobility systems are increasingly dependent on precise and scalable simulation platforms to facilitate the design, assessment, and optimization of intelligent transportation systems (ITS). This paper sets forth a thorough benchmarking investigation of three extensively utilized traffic simulation platforms: SUMO (Simulation of Urban MObility), CityFlow, and MATSim, with an emphasis on their computational efficacy within large-scale synthetic traffic scenarios. We conduct an evaluation of these platforms across vehicle volumes ranging from 10 to 10 million vehicles, assessing total runtime, CPU and GPU utilization, and memory consumption on high-performance computing infrastructures. Our results reveal notable architectural trade-offs: SUMO exhibits predictable linear scaling but becomes constrained by CPU limitations at elevated vehicle counts, CityFlow encounters memory limitations beyond 10,000 vehicles, whereas MATSim necessitates meticulous JVM (Java Virtual Machine) optimization to effectively manage large-scale scenarios. This investigation provides essential guidance for researchers and practitioners in the selection of suitable simulation tools for urban-scale traffic modeling and identifies critical computational challenges associated with the scaling of simulations for smart city initiatives., This work was supported by the grant PID2023-148716OB-C31 funded by MCIU/AEI/10.13039/501100011033 (DISCOVERY project). Also, by the project ”MultiMO: Datos MultiSectoriales para la Movilidad Obligada” TSI100123-2024-60 (European Union, NextGenerationEU); Also, by predoctoral scholarship for the training of research personnel associated with the ”Generaci´on de Conocimiento” Project PRE2021-099830; Finally, by the Generalitat de Catalunya under AGAUR grant ”2021 SGR 01413”., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

MobilitApp: a deep learning-based tool for transport mode detection to support sustainable urban mobility

  • Caravaca Ibáñez, Gerard
  • Cruz Llopis, Luis Javier de la
  • Catalin Diaconeasa, Adrian
  • Bazán Guillén, Alberto
  • Aguilar Igartua, Mónica
The visible effects of climate change in urban areas are driving a paradigm shift in societal and political priorities. Urban planners, public transport providers, and traffic managers are increasingly focused on redesigning cities to promote sustainable mobility and create green spaces for pedestrians, cyclists, and scooter users. In alignment with these objectives, the European Climate Law mandates a minimum 55% reduction in greenhouse gas emissions by 2030 and climate neutrality by 2050. Achieving these targets requires robust tools to collect and analyze mobility data, enabling the evaluation of citizens’ travel habits and the planning of sustainable urban infrastructure. This study presents MobilitApp, a tool developed based on a deep learning (DL) model for real-time detection of transportation modes using smartphone sensor data. Our approach leverages a hierarchical model combining convolutional neural networks (CNNs) for feature extraction and long short-term memory (LSTM) layers for temporal processing, enhanced by skip connections. To ensure computational efficiency on mobile devices, the system integrates statistical techniques for early motion detection, minimizing reliance on DL models. The model was trained on a dataset of multimodal trips in Barcelona, achieving over 80% accuracy for most transport modes and a weighted average accuracy of 88%. These results highlight the effectiveness of our approach for accurately predicting users’ transport modes during their trips. The MobilitApp tool provides an intuitive platform for collecting and analyzing urban mobility data. By analyzing travel patterns, transport modes, and mode-switching behaviors, it delivers actionable insights to city planners, aiding in the enhancement of urban mobility, promotion of sustainable development, and transition to greener cities., This work was supported in part by Spanish Government funded by MCIN/AEI/10.13039/501100011033 under Research Projects ’’DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)’’ under Grant PID2023-148716OB-C32 and ’’Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)’’ under Grant PID2020-113795RB-C31; partially funded by MCIN/AEI/10.13039/501100011033 and by the European Union (EU) NextGenerationEU/PRTR (Plan de Recuperación, Transformación y Resiliencia) under Research Project ’’Anonymization Technology for AI-Based Analytics of Mobility Data (MOBILYTICS)’’ under Grant TED2021-129782B-I00; in part by the Predoctoral Scholarship ‘‘Generación de Conocimiento- Projects Call 2022’’ under Grant PRE2021-099830; and in part by the Generalitat de Catalunya under AGAUR(Agència de Gestió d’Ajuts Universitaris i de Recerca) Grant 2021-SGR-01413., Peer Reviewed, Objectius de Desenvolupament Sostenible::11 - Ciutats i Comunitats Sostenibles, Objectius de Desenvolupament Sostenible::11 - Ciutats i Comunitats Sostenibles::11.2 - Per a 2030, proporcionar accés a sistemes de transport segurs, assequibles, accessi­bles i sostenibles per a totes les persones, i millorar la seguretat viària, en particular mitjan­çant l’ampliació del transport públic, amb especial atenció a les necessitats de les persones en situació vulnerable, dones, nenes, nens, persones amb discapacitat i persones grans, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

On the adaptation of trajectory generative models to enforce differential privacy through gradient perturbation

  • Rubio Jornet, Víctor
  • Parra Arnau, Javier
  • Forné Muñoz, Jorge
Synthetic trajectories have become a crucial tool for enabling data sharing in domains such as telecommunications and urban planning, yet most generators provide neither formal privacy guarantees nor empirical validation of the protection they claim. This paper investigates whether state-of-the-art deep learning trajectory generators can be adapted to satisfy differential privacy (DP) while retaining usable utility, and makes three contributions. First, we analyze 16 state-of-the-art generators and determine which can be trained with Differentially Private Stochastic Gradient Descent (DP-SGD), showing that common architectural choices — such as index-based coordinate look-ups at inference — silently void the guarantee. Second, we present Sec-LSTM-TrajGAN, a DP adaptation of LSTM-TrajGAN, a well-known trajectory generator chosen after the analysis of the existing models. Third, we introduce RouteMIA, to our knowledge the first closed-box membership inference attack tailored to synthetic trajectories, which adapts the distanceto-closest-record attack to sequential spatio-temporal data and requires access only to the released dataset. Evaluating on GeoLife (17, 000 trajectories) and Porto (100, 000 trajectories) datasets for several scenarios of increasing privacy. RouteMIA attains an AUC of 0.59 and 0.55 against the non-private models, which drops to ˜ 0.50 — chance level — for every private configuration. Notably, gradient clipping alone, with no noise added, already accounts for this entire empirical gain, at a utility cost of a distance JSD rising from 0.12 to 0.20 on Porto and essentially none on GeoLife, whereas strict budgets (e = 3) degrade GeoLife markedly (distance JSD up to 0.35). Per-example clipping, not the noise, therefore delivers the measurable protection against this attack — a cheap empirical complement to, but not a replacement for, formal guarantees., This work was supported in part by Spanish Government, funded by Ministerio de Ciencia e Innovación (MCIN)/Agencia Estatal de Investigación (AEI)/10.13039/501100011033 through the research project ‘‘DIstributed Smart COmmunications with Verifiable EneRgy-optimal Yields (DISCOVERY)’’ under Grant PID2023-148716OB-C32, through the research project ‘‘Privacy-Preserving Techniques for Mobility Data (PRIMOD)’’ under Grant CNS2025-166870, through the research project ‘‘Anonymization Technology for Artificial Intelligence (AI)-Based Analytics of Mobility Data (MOBILYTICS)’’ under Grant TED2021-129782B-I00, and through the research project ‘‘Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)’’ under Grant PID2020-113795RB-C31; in part by the project MOBILYTICS funded by European Union—NextGenerationEU/Recovery, Transformation and Resilience Plan (PRTR); and in part by the Generalitat de Catalunya through AGAUR under Grant 2021 SGR 01413. The work of Víctor Rubio-Jornet was supported by the Joan Oró Agència de Gestió d’Ajuts Universitaris i de Recerca (AGAUR)-Formació d’Investigadors (FI) Predoctoral Program, funded by the Secretariat of Universities and Research of the Department of Research and Universities of the Generalitat of Catalonia and European Social Fund Plus under Grant 2024 FI-1 00762. The work of Javier Parra-Arnau was supported by Ramón y Cajal Grant through Spanish Ministry of Science and Innovation and European Union—‘‘NextGenerationEU’’/PRTR under Grant RYC2021-034256-I., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

RUTGe: realistic urban traffic generator for urban environments using deep reinforcement learning and SUMO simulator

  • Bazán Guillén, Alberto
  • Barbecho Bautista, Pablo
  • Aguilar Igartua, Mónica
We are witnessing a profound shift in societal and political attitudes, driven by the visible consequences of climate change in urban environments. Urban planners, public transport providers, and traffic managers are urgently reimagining cities to promote sustainable mobility and expand green spaces for pedestrians, bicycles, and scooters. To design more sustainable cities, urban planners require realistic simulation tools to optimize mobility, identify location for car chargers, convert streets to pedestrian zones, and evaluate the impact of alternative configurations. However, realistic traffic profiles are essential to produce meaningful simulation results. Addressing this need, we propose a traffic generator based on deep reinforcement learning integrated with the SUMO simulator. This tool learns to generate an instantaneous number of vehicles throughout the day, aligning closely with the target profiles observed at the traffic monitoring stations. Our approach generates accurate 24-hour traffic patterns for any city using minimal statistical data, achieving higher accuracy compared to existing alternatives. In particular, our proposal demonstrates a highly accurate 24-hour traffic adjustment, with the generated traffic deviating only by about 5% from the real target traffic. This performance significantly exceeds that of current SUMO tools like RouteSampler, which struggle to accurately follow the total daily traffic curve, especially during peak hours when severe traffic congestion occurs., This work was partially supported by the Spanish Government under these research projects funded by MCIN/AEI/10.13039/501100011033: DISCOVERY PID2023-148716OB-C32; MOBILYTICS TED2021-129782B-I00 (also funded by the European Union NextGenerationEU/PRTR); COMPROMISE PID2020-113795RB-C31; predoctoral scholarship associated with the ”Generaci´ on de Conocimiento” Projects, Call 2022, PRE2021099830. Also, by the Generalitat de Catalunya AGAURgrant ”2021 SGR 01413”., Peer Reviewed, 11 - Ciutats i Comunitats Sostenibles, 11.2 - Per a 2030, proporcionar accés a sistemes de transport segurs, assequibles, accessibles i sostenibles per a totes les persones, i millorar la seguretat viària, en particular mitjançant l’ampliació del transport públic, amb especial atenció a les necessitats de les persones en situació vulnerable, dones, nenes, nens, persones amb discapacitat i persones grans, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Uncoordinated syntactic privacy: a new composable metric for multiple, independent data publishing

  • Tobar Nicolau, Adrián
  • Parra Arnau, Javier
  • Forné Muñoz, Jorge
  • Torra, Vicenç
A privacy model is a privacy condition, dependent on a parameter, that guarantees an upper bound on the risk of reidentification disclosure and maybe also on the risk of attribute disclosure by an adversary. A privacy model is composable if the privacy guarantees of the model are preserved, possibly to a limited extent, after repeated independent application of the privacy model. From the opposite perspective, a privacy model is not composable if multiple independent data releases, each of them satisfying the requirements of the privacy model, may result in a privacy breach. Current privacy models are broadly classified into syntactic ones (such as k-anonymity and l-diversity) and semantic ones, which essentially refer to e -differential privacy (e-DP) and variations thereof. While e-DP and its variants offer strong composability properties, syntactic notions are not composable unless data releases are conducted by a single, centralized data holder that uses specialized notions such as m-invariance and t -safety. In this work, we propose m-uncoordinated-syntactic-privacy (m-USP), the first syntactic notion with composability properties for the independent publication of nondisjoint data, in other words, without a centralized data holder. Theoretical results are formally proven, and experimental results demonstrate that the risk to individuals does not increase significantly, in contrast to non-composable methods, that are susceptible to attribute disclosure. In most cases, the utility degradation caused by the extra protection is less than 5% and decreases as the value of m increases., Javier Parra-Arnau is the recipient of a “Ram´ on y Cajal” fellowship (ref. RYC2021-034256-I) funded by the Spanish Ministry of Science and Innovation and the European Union– “NextGenerationEU”/PRTR (Plan de Recuperaci´ on, Transformaci´ on y Resiliencia). This work was also supported by the Spanish Government under the following projects: ”DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)” PID2023-148716OB-C32, funded by MCIN/AEI/10.13039/501100011033; “Enhancing Communication Protocols with Machine Learning while Protecting Sensitive Data (COMPROMISE)” PID2020-113795RBC31, funded by MICIU/AEI/10.13039/501100011033; and “Anonymization technology for AI-based analytics of mobility data (MOBILYTICS)” (TED2021-129782B-I00), funded by MICIU/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR. This work was also partially funded by the Generalitat de Catalunya, under AGAUR grant “2021 SGR 01413”; was partially supported by the Wallenberg AI, Autonomous Systems and Software Program (WASP) funded by the Knut and Alice Wallenberg Foundation; and received support by the project ”Privacy for complex data” (VR 2022-04645)., Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

FedITS-tool: A lockstep co-simulation framework for vehicular federated learning with per-round energy and carbon accounting

  • Junjun, Lu
  • Postigo Boix, Marcos
  • Bazán Guillén, Alberto
  • Cruz Llopis, Luis Javier de la
  • Aguilar Igartua, Mónica
Vehicular federated learning (VFL) under realistic mobility should be evaluated not only by model accuracy, but also by the energy and carbon cost of distributed training. Mobility, intermittent connectivity, and deadline misses can cause clients to consume energy for updates that are later dropped before aggregation, producing wasted carbon emissions. This paper presents FedITS-Tool, a modular lockstep co-simulation framework that couples SUMO-based vehicle mobility, a lightweight Veins/OMNeT++ connectivity-time service, and Flower-based FL training in Docker containers. A central orchestrator synchronizes simulation time with FL rounds, maps available clients to simulated vehicles, enforces deadline-aware commit/drop semantics, and exports per-round traces of failure causes, communication delays, energy use, and carbon emissions. Experiments show that FedITS-Tool quantifies up to 40% dropped-carbon waste in coverage-stressed scenarios while preserving traceability in 500-vehicle simulations. The tool supports reproducible studies of RSU coverage planning, deadline analysis, participation dynamics, and carbon-aware VFL design., This work was supported by the project “DISCOVERY: Distributed Smart Communications with Verifiable Energy-optimal Yields” PID2023-148716OB-C32 (Agencia Estatal de Investigación, Ministerio de Ciencia e Innovación); also by the project “MultiMO: Datos MultiSectoriales para la Movilidad Obligada” TSI-100123-2024-60 (Ministerio de transformación Digital y de la función pública, NextGenerationEU); also, by a predoctoral scholarship for the training of research personnel associated with the “Generación de Conocimiento” Project PRE2021-099830., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Comparing optimal and adaptive EV charging in smart cities: MILP vs. reinforcement learning

  • Bazán Guillén, Alberto
  • Barbecho Bautista, Pablo
  • Aguilar Igartua, Mónica
  • Cuomo, Francesca
The coordinated scheduling of electric vehicle (EV) charging is a critical challenge for smart cities, particularly in high-density infrastructure such as Mobility Hubs (MHs). This paper evaluates and compares two prominent approaches to the EV Charging Scheduling Problem (CSP): Mixed-Integer Linear Programming (MILP) and Reinforcement Learning (RL). We formulate a shared problem framework and apply both strategies under two structured scenarios: a small-scale deterministic benchmark and a medium-scale, realistic deployment with higher heterogeneity. Results show that MILP achieves optimal cost and state of charge SoC compliance in tractable cases but struggles with scalability. RL, based on Proximal Policy Optimization (PPO), achieves near-optimal performance while scaling to 100 EVs with minimal computation time. Despite occasional SoC deviations, the RL agent exhibits robust and adaptive behavior under dynamic conditions. This study offers actionable insights for selecting and deploying EV scheduling strategies in real-world urban environments., This work was partially supported by the Spanish Government under this research project funded by MCIN/AEI/10.13039/501100011033: DISCOVERY PID2023-148716OB-C32. Also by the project MultiMO TSI100123-2024-0060; and by the predoctoral scholarship associated with the ”Generaci´on de Conocimiento” Projects, Call 2022, PRE2021-099830. Also, by the Generalitat de Catalunya AGAUR grant ”2021 SGR 01413”., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

An AI-based intelligent vehicle routing approach for large-scale fleet optimization

  • Guin, Agneev
  • Aguilar Igartua, Mónica
This research addresses the challenges of urban mobility and logistics by developing an AI-based system for intelligent route prediction and coordination. Using Graph Neural Networks and complementary AI methods, we intend to model heterogeneous fleets to optimize large-scale mobility with real-time decision-making and energy-aware trade-offs., This work was supported by the DISCOVERY project PID2023-148716OBC31 funded by MCIU/AEI/10.13039/501100011033. Also, by the MultiMO project TSI-100123-2024-60. Finally, by the Generalitat de Catalunya under AGAUR grant 2021 SGR 01413., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Deployment-aware multi-objective optimization of machine learning models for predictive congestion control in multi-hop vehicular networks

  • Astudillo León, Juan Pablo
  • Lemus Cárdenas, Leticia
  • Cruz Llopis, Luis Javier de la
  • Aguilar Igartua, Mónica
Multi-hop vehicular wireless networks are highly susceptible to channel congestion due to dynamic topologies, heterogeneous traffic densities, and contention-based medium access. This paper proposes a deployment-aware predictive congestion control framework for vehicular communications, where packet transmission decisions are guided by supervised machine learning models optimized under predictive and computational constraints. Its novelty lies in coupling packet-level congestion prediction with deployment-aware model selection, allowing transmission-control models to be selected according to predictive quality, measured native-inference latency, and serialized model footprint. A packet-level dataset was constructed from realistic vehicle trajectories generated with SUMO using road networks extracted from OpenStreetMap for three topologically diverse cities: Liverpool, Rio de Janeiro, and Houston. These trajectories support vehicular network evaluation under heterogeneous urban conditions, enabling the assessment of predictive transmission decisions across different road topologies. Several supervised learning models were considered, including linear classifiers, neural networks, and ensemble machine learning architectures such as Random Forest, Weighted Soft Voting, and Stacking. Hyperparameter tuning was formulated as a multi-objective optimization problem and solved using the NSGA-II evolutionary algorithm, jointly maximizing predictive performance while minimizing inference latency and serialized model size. Pareto-efficient models were exported to ONNX format and executed through a native inference pipeline to evaluate their suitability for computationally constrained vehicular communication environments. Results across heterogeneous urban scenarios show that the proposed framework improves vehicular communication performance in terms of packet delivery reliability and communication delay, while preserving competitive machine learning performance under deployment-oriented constraints., This work was supported in part by Spanish Government funded by MCIN/AEI/10.13039/501100011033 under Research Project ‘‘DIstributed Smart Communications with Verifiable EneRgy-optimal Yields (DISCOVERY)’’ under Grant PID2023-148716OB-C32. The work of Juan Pablo Astudillo León was supported by Yachay Tech University for the support given to the ‘‘VICOM-AI: Vehicular Intelligent Communications Optimization Using Artificial Intelligence Project’’ under Grant MATH25-05., Peer Reviewed, Postprint (published version)