Participation in Research Projects
Participation in Research Projects
Embedded Files
European projects:
European projects:
H2020 Captor Project [2016-2019]:
National projects:
National projects:
PID2022-138155OB-I00 [2023-2026]. DIDEROT: “Data-driven methods to enhance the information quality in IoT networks”.
PID2019-107910RB-I00 [2020-2022]: IoT monitoring of air quality (IMAQ), January 2020 - December 2022
TIN2016-78473-C3-1-R [2017-2019] : Semantic IoT Data Integration and its use in the energy management of Smart Grids and Smart Cities (Integración Semántica de datos de IoT y su uso en la gestión energética de las Smart Grids y Smart Cities), SEMIOTIC, January 2017 - December 2019.
Regional projects:
Regional projects:
CLIMA 00097 [2024-2025]. URBANAT: "Under the skin of the city: Urban simulations for nature-based solutions".
AGAUR 2017SGR-990 [2017-2019]: Computer Networks and Distributed Systems (Regional Project), January 2017 - December 2019
AGAUR 2021SGR
Company-Related projects:
Company-Related projects:
CDTI QCDI (Quantum Cognitive Digital Industry). Company: Repsol.
PERTE DICAROS. Company: AKO Electromecànica.
SENSGEM. Company: Mensoft Consultores.
Calibration of Low-Cost Air Pollution Sensors
Calibration of Low-Cost Air Pollution Sensors
Low-cost sensors present a great alternative for measuring air quality, complementing the high-precision instrumentation used by governments. However, these sensors present data quality issues that need to be addressed in order to be used as reliably as possible.
Low-cost sensors present a great alternative for measuring air quality, complementing the high-precision instrumentation used by governments. However, these sensors present data quality issues that need to be addressed in order to be used as reliably as possible.
Low-Cost Sensor Pre-processing
Low-Cost Sensor Pre-processing
Before air quality measurements can be obtained, several steps are essential; sensor sampling, data filtering, and data aggregations. Energy-efficient sampling techniques are of special interest for battery-powered Internet of Things nodes, including duty cycle strategies.
Machine Learning-Based in-situ Sensor Calibration
Machine Learning-Based in-situ Sensor Calibration
Once sensor measurements are available, a mechanism is needed to correct the quality of the sensor data and translate from raw sensor values to pollutant concentrations. Supervised machine learning techniques (e.g., SVR, RF, MLR, ANN) are used for this purpose.
Machine learning is widely used in the field of low-cost sensors for air quality monitoring. For in-situ calibration, a low-cost sensor is placed next to a precision instrument to train a supervised machine learning model, such as multiple linear regression, support vector regression or neural networks. There is also the multisensor approach, where measurements from different sensors measuring the same pollutant are combined to increase the accuracy of the calibration, as well as to provide more reliability and resilience. More machine learning techniques have been used for air pollution forecasting using low-cost sensors.
Machine learning is widely used in the field of low-cost sensors for air quality monitoring. For in-situ calibration, a low-cost sensor is placed next to a precision instrument to train a supervised machine learning model, such as multiple linear regression, support vector regression or neural networks. There is also the multisensor approach, where measurements from different sensors measuring the same pollutant are combined to increase the accuracy of the calibration, as well as to provide more reliability and resilience. More machine learning techniques have been used for air pollution forecasting using low-cost sensors.
For further information, refer to:
For further information, refer to:
P. Ferrer-Cid, J. Garcia-Calvete, A. Main-Nadal, Z. Ye, J. M. Barcelo-Ordinas, J. Garcia-Vidal Sampling Trade-Offs in Duty-Cycled Systems for Air Quality Low-Cost Sensors https://doi.org/10.3390/s22103964, Sensors 2022, 22(10), 3964.
P. Ferrer-Cid, J. Garcia-Calvete, A. Main-Nadal, Z. Ye, J. M. Barcelo-Ordinas, J. Garcia-Vidal Sampling Trade-Offs in Duty-Cycled Systems for Air Quality Low-Cost Sensors https://doi.org/10.3390/s22103964, Sensors 2022, 22(10), 3964.
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal, A. Ripoll, M. Viana, Multi-sensor Data Fusion Calibration in IoT Air Pollution Platforms, https://doi.org/10.1109/JIOT.2020.2965283, IEEE Internet of Things Journal (IEEE IoT-J), 2020
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal, A. Ripoll, M. Viana, Multi-sensor Data Fusion Calibration in IoT Air Pollution Platforms, https://doi.org/10.1109/JIOT.2020.2965283, IEEE Internet of Things Journal (IEEE IoT-J), 2020
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal, A. Ripoll, M. Viana, A Comparative Study of Calibration Methods for Low-Cost Ozone Sensors in IoT Platforms, https://doi.org/10.1109/JIOT.2019.2929594, IEEE Internet of Things Journal (IEEE IoT-J), 2019
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal, A. Ripoll, M. Viana, A Comparative Study of Calibration Methods for Low-Cost Ozone Sensors in IoT Platforms, https://doi.org/10.1109/JIOT.2019.2929594, IEEE Internet of Things Journal (IEEE IoT-J), 2019
Barcelo-Ordinas, J.M.; Ferrer-Cid, P.; Garcia-Vidal, J.; Ripoll, A.; Viana, M. Distributed Multi-Scale Calibration of Low-Cost Ozone Sensors in Wireless Sensor Networks. https://doi.org/10.3390/s19112503 Sensors 2019, 19, 2503.
Barcelo-Ordinas, J.M.; Ferrer-Cid, P.; Garcia-Vidal, J.; Ripoll, A.; Viana, M. Distributed Multi-Scale Calibration of Low-Cost Ozone Sensors in Wireless Sensor Networks. https://doi.org/10.3390/s19112503 Sensors 2019, 19, 2503.
Graph-Based Analysis of Air Pollution Sensor Networks
Graph-Based Analysis of Air Pollution Sensor Networks
Once a sensor network is deployed to monitor air quality in an area of interest it is of great importance to maintain the quality of the network data. The recent graph signal processing framework allows to analyze signals defined on graphs, thus allowing to translate classical signal processing techniques to the irregular domain of graphs. Therefore, using a graph-based approach, it is possible to describe the relationships existing in a sensor network, containing both low-cost sensors and high-precision sensors, in order to be able to overlay machine learning techniques to maintain data quality.
Once a sensor network is deployed to monitor air quality in an area of interest it is of great importance to maintain the quality of the network data. The recent graph signal processing framework allows to analyze signals defined on graphs, thus allowing to translate classical signal processing techniques to the irregular domain of graphs. Therefore, using a graph-based approach, it is possible to describe the relationships existing in a sensor network, containing both low-cost sensors and high-precision sensors, in order to be able to overlay machine learning techniques to maintain data quality.
Graph-Based Sensor Network Representation
Graph-Based Sensor Network Representation
In order to carry out a graph-based analysis, it is first necessary to find the graph that best represents the relationships between the different sensors in the network. Examples of graph learning techniques include; graphical Lasso, distance-based graph, and smoothness-based graph.
Machine Learning Applied Over Graphs
Machine Learning Applied Over Graphs
Once a graph describing the measurements of a sensor network is available, graph signal processing and machine learning techniques can be superimposed to perform different applications (e.g., signal filtering, clustering, data reconstruction).
The topology describing a sensor network can be used in conjunction with machine learning and signal processing techniques to carry out different applications. Among the applications, we highlight the use of signal filtering, node clustering, and data reconstruction. Data reconstruction can be used to perform real-time tasks such as missing value imputation and virtual sensing to maintain data quality in a network where different low-cost sensors can suffer failures due to communication failures, sensor malfunction, and others.
The topology describing a sensor network can be used in conjunction with machine learning and signal processing techniques to carry out different applications. Among the applications, we highlight the use of signal filtering, node clustering, and data reconstruction. Data reconstruction can be used to perform real-time tasks such as missing value imputation and virtual sensing to maintain data quality in a network where different low-cost sensors can suffer failures due to communication failures, sensor malfunction, and others.
For further information, refer to:
For further information, refer to:
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal Graph Signal Reconstruction Techniques for IoT Air Pollution Monitoring Platforms https://doi.org/10.1109/JIOT.2022.3196154 IEEE Internet of Things Journal (IEEE IoT-J), 2022
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal Graph Signal Reconstruction Techniques for IoT Air Pollution Monitoring Platforms https://doi.org/10.1109/JIOT.2022.3196154 IEEE Internet of Things Journal (IEEE IoT-J), 2022
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal Data Reconstruction Applications for IoT Air Pollution Sensor Networks using Graph Signal Processing https://doi.org/10.1016/j.jnca.2022.103434, Elsevier Journal of Network and Computer Applications (JNCA), 2022
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal Data Reconstruction Applications for IoT Air Pollution Sensor Networks using Graph Signal Processing https://doi.org/10.1016/j.jnca.2022.103434, Elsevier Journal of Network and Computer Applications (JNCA), 2022
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal Graph Learning Techniques Using Structured Data for IoT Air Pollution Monitoring Platforms https://doi.org/10.1109/JIOT.2021.3067717, IEEE Internet of Things Journal (IEEE IoT-J), 2021
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal Graph Learning Techniques Using Structured Data for IoT Air Pollution Monitoring Platforms https://doi.org/10.1109/JIOT.2021.3067717, IEEE Internet of Things Journal (IEEE IoT-J), 2021
Graph-Based Anomaly Detection
Graph-Based Anomaly Detection
Since data quality is one of the most important aspects of low-cost sensors, the detection of outliers and faulty sensors are of particular importance in this field. Thus, the topology of the network together with the information from the sensors can be of help to detect and locate if any sensor provides anomalous measurements.
Since data quality is one of the most important aspects of low-cost sensors, the detection of outliers and faulty sensors are of particular importance in this field. Thus, the topology of the network together with the information from the sensors can be of help to detect and locate if any sensor provides anomalous measurements.
Volterra Graph-Based Outlier Detection
Volterra Graph-Based Outlier Detection
An algorithm combining a graph learned from the network data and a graph signal reconstruction model based on the Volterra series can be used to jointly detect outliers. Hence, an adaptive algorithm based on signal reconstruction residuals' thresholding can be very useful in the detction of faulty measurements and sensors.
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal Volterra Graph-Based Outlier Detection for Air Pollution Sensor Networks https://doi.org/10.1109/TNSE.2022.3169220, IEEE Transactions on Network Science and Engineering (IEEE TNSE), 2022
P. Ferrer-Cid, J. M. Barcelo-Ordinas, J. Garcia-Vidal Volterra Graph-Based Outlier Detection for Air Pollution Sensor Networks https://doi.org/10.1109/TNSE.2022.3169220, IEEE Transactions on Network Science and Engineering (IEEE TNSE), 2022
Page updated
Google Sites
Report abuse