Outdoor NB-IoT and 5G coverage and channel information data in urban environments

This dataset includes data for NB-IoT and 5G networks as collected in two cities: Oslo, Norway (NB-IoT only) and Rome, Italy (both NB-IoT and 5G).

Data were collected using the Rohde & Schwarz TSMA6 mobile network scanner. 7 measurement campaigns are provided for Oslo, and 6 for Rome. Additional data collected in Rome are provided in the following large-scale dataset, focusing on the two major mobile network operators: https://ieee-dataport.org/documents/large-scale-dataset-4g-nb-iot-and-5g-non-standalone-network-measurements 

The dataset includes a metadata file providing the following information for each campaign: 

date of collection;

start time and end time of collection;

length;

type (walking/driving).

Two additional metadata files are provided: two .kml files, one for each city, allowing the import of coordinates of data points organized by campaign in a GIS engine, such as Google Earth, for interactive visualization.

The dataset contains the following data for NB-IoT:

Raw data for each campaign, stored in two .csv files. For a generic campaign , the files are:

NB-IoT_coverage_C.csv including a geo-tagged data entry in each row. Each entry provides information on a Narrowband Physical Cell Identifier (NPCI), with data related to the time stamp the NPCI was detected, GPS information, network (NPCI, Operator, Country Code, eNodeB-ID) and RF signal (RSSI, SINR, RSRP and RSRQ values);

 NB-IoT_RefSig_cir_C.csv, also including a geo-tagged data entry in each row. Each entry provides information on a NPCI, with data related to the time stamp the NPCI was detected, GPS information, network (NPCI, Operator ID, Country Code, eNodeB-ID) and Channel Impulse Response (CIR) statistics, including the maximum delay.

Processed data, stored in a Matlab workspace (.mat) file for each city:  data are grouped in data points, identified by pairs. Each data point provides RF and CIR maximum delay measurements for each unique combination detected at the coordinates of the data point.

Estimated positions of eNodeBs, stored in a csv file for each city;

A matlab script and a function to extract and generate processed data from the raw data for each city.

The dataset contains the following data for 5G:

Raw data for each campaign, stored in two .xslx files. For a generic campaign , the files are:

5G_coverage_C.xslx including a geo-tagged data entry in each row. Each entry provides information on a Physical Cell Identifier (PCI), with data related to the time stamp the PCI was detected, GPS information, network (PCI, Beamforming Index, Operator, Country Code) and RF data (SSB-RSSI, SSS-SINR, SSS-RSRP and SSS-RSRQ values, and similar information for the PBCH signal);

 5G_RefSig_cir_C.csv, also including a geo-tagged data entry in each row. Each entry provides information on a PCI, with data related to the time stamp the PCI was detected, GPS information, network (PCI, Beamforming Index, Operator ID, Country Code) and Channel Impulse Response (CIR) statistics, including the maximum delay.

Processed data, stored in a Matlab workspace (.mat) file:  data are grouped in data points, identified by pairs. Each data point provides RF and CIR maximum delay measurements for each unique combination detected at the coordinates of the data point.

A matlab script and a supporting function to extract and generate processed data from the raw data.

In addition, in the case of the Rome data additional matlab workspaces are provided, containing interpolated data in the feature dimensions according to two different approaches:

A campaign-by-campaign linear interpolation (both NB-IoT and 5G);

A bidimensional interpolation on all campaigns combined (NB-IoT only).

A function to interpolate missing data in the original data according to the first approach is also provided for each technology. The interpolation rationale and procedure for the first approach is detailed in:

L. De Nardis, G. Caso, Ö. Alay, U. Ali, M. Neri, A. Brunstrom and M.-G. Di Benedetto, "Positioning by Multicell Fingerprinting in Urban NB-IoT networks," Sensors, Volume 23, Issue 9, Article ID 4266, April 2023. DOI: 10.3390/s23094266.

The second interpolation approach is instead introduced and described in:

L. De Nardis, M. Savelli, G. Caso, F. Ferretti, L. Tonelli, N. Bouzar, A. Brunstrom, O. Alay, M. Neri, F. Elbahhar and M.-G. Di Benedetto, " Range-free Positioning in NB-IoT Networks by Machine Learning: beyond WkNN", under major revision in IEEE Journal of Indoor and Seamless Positioning and Navigation.

Positioning using the 5G data was furthermore in investigated in: 

K. Kousias, M. Rajiullah, G. Caso, U. Ali, Ö. Alay, A. Brunstrom, L. De Nardis, M. Neri, and M.-G. Di Benedetto, "A Large-Scale Dataset of 4G, NB-IoT, and 5G Non-Standalone Network Measurements," IEEE Communications Magazine, Volume 62, Issue 5, pp. 44-49, May 2024. DOI:  10.1109/MCOM.011.2200707.

G. Caso, M. Rajiullah, K. Kousias, U. Ali, N. Bouzar, L. De Nardis, A. Brunstrom, Ö. Alay, M. Neri and M.-G. Di Benedetto,"The Chronicles of 5G Non-Standalone: An Empirical Analysis of Performance and Service Evolution", IEEE Open Journal of the Communications Society, Volume 5, pp. 7380 - 7399, 2024. DOI: 10.1109/OJCOMS.2024.3499370.

Please refer to the above publications when using and citing the dataset.