Narrowband Internet of Things (NB-IoT) has quickly become a leading technology in the deployment of IoT systems and services, thanks to its appealing features in terms of coverage, energy efficiency, and compatibility with existing mobile networks. Increasingly, IoT services and applications require location information to be paired with data collected by devices. NB-IoT does not provide, however, reliable information on positioning. Time-based techniques inherited from Long Term Evolution (LTE), such as Observed Time Difference Of Arrival (OTDOA) , are not yet widely available in existing networks, and are expected to perform poorly on NB-IoT signals due to their narrow bandwidth.
The goal of the project is to introduce a suite of algorithms for NB-IoT positioning based on fingerprinting, as very recently proposed for NB-IoT networks. The algorithms will combine network coverage, radio frequency (RF) signal parameters and Time of Arrival (ToA) information, collected from multiple cells. Multiple levels of hybridization between RF and ToA will be investigated, ranging from a low-level integration, with a single fingerprint including both RF and ToA data, to a high-level integration, with RF and ToA used to obtain two independent position estimates that are then averaged. The proposed strategies will be evaluated on experimental data for multiple NB-IoT operators, obtained in two large-scale urban measurement campaigns performed in Oslo, Norway, and Rome, Italy. Data will be preprocessed in order to combine RF and ToA information in each data measurement location, enabling the definition of hybrid fingerprints, and the resulting processed dataset will be released under an open-source license. Performance evaluation will also include a comparison between the proposed algorithms and OTDOA, carried out on both NB-IoT and LTE data (also part of the experimental data).
The project will go beyond the current state of the art in the field of positioning in cellular networks under the following aspects:
- It will investigate the applicability of positioning by fingerprinting to NB-IoT, beyond the early results available in the literature today, restricted to a rather simple definition of fingerprint based on RSSI collected in a single cell, and to the use of the WkNN algorithm;
- It will introduce new definitions of fingerprints, combining coverage, RF and ToA information, beyond the current proposals in 3GPP technical specifications for LTE, limited at best to a straightforward normalized sum of a single RF parameter and the ToA information;
- It will introduce new similarity metrics. The choice of the metric to determine the similarity between two fingerprints (or, conversely their distance) is a key aspect in the design of a fingerprinting system. As a consequence, a plethora of metrics have been proposed in the literature; however, they typically focused on the simple case of a monodimensional fingerprint including only one value for each detected base station. This project will instead define similarity metrics that consider the presence of heterogeneous data in a single fingerprint, typically taking value over different ranges, and of multiple values being associated to the same NPCI in a fingerprint.
- It will propose new fingerprinting algorithms that take advantage of the combination of RF and ToA information at different levels, ranging from full information merging (fingerprint definition combining the two information pieces) to high level combination (weighted average of the position estimates obtained independently using RF and ToA);
- It will provide the first comprehensive assessment of positioning accuracy in NB-IoT networks in large urban scenarios using real world data, covering both fingerprinting and ToA-based approaches, whereas performance evaluation carried out in existing investigations was typically based on system simulations;
- Even for positioning in LTE, that will be studied in order to highlight the differences between this technology and NB-IoT, the project will provide a comparison between OTDOA and fingerprinting based on large scale real world data that is currently not available in the literature: the original proposal of fingerprinting for LTE released by 3GPP, for example, was only tested by simulations;
- It will provide a reliable comparison of the performance of the positioning algorithms when used with different Radio Access Technologies (NB-IoT vs. LTE) and in networks deployed and run by different operators. The capability of the Rohde&Schwarz hardware to collect data simultaneously for all operators and for multiple technologies will allow to isolate two key aspects that affect the performance of a network in general, and of the positioning service running on it: 1) the network deployment adopted by the operator and 2) the selected RAT. The analysis will go beyond the current state of the art, because the availability of the data for the different RATs and operators in the same exact set of locations will allow to remove the bias due to different data collection setups typically present when using data for different RATs and different operators.
- It will make available under an open-source license a preprocessed dataset combining RF and ToA information collected over several thousands of locations in a urban scenario, beyond anything currently available for research purposes. The dataset will be the result of the preprocessing activity planned in the project and will ensure coherency between RF and ToA information, by providing the information already combined at each location. This will enable researchers to extend the results obtained in the project, by designing new positioning algorithms combining the two information pieces, without the need for any time-consuming data preprocessing. The dataset will also include the estimates of the positions of the eNBs obtained using both the built-in mapping capability of the Rohde&Schwarz hardware and the position estimation algorithms developed during the preprocessing activity. The availability in the dataset of both ToA and positions of the base stations will open the way to the design and evaluation of new ranging-based positioning schemes as well.