Crowdsourced Reconstruction of Cellular Networks to Serve Outdoor Positioning: Modeling, Validation and Analysis

Positioning via outdoor fingerprinting, which exploits the radio signals emitted by cellular towers, is fundamental in many applications. In most cases, the localization performance is affected by the availability of information about the emitters, such as their coverage. While several projects aim at collecting cellular network data via crowdsourcing observations, none focuses on information about the structure of the networks, which is paramount to correctly model their topology. The difficulty of such a modeling is exacerbated by the inherent differences among cellular technologies, the strong spatio-temporal nature of positioning, and the continuously evolving configuration of the networks. In this paper, we first show how to synthesize a detailed conceptual schema of cellular networks on the basis of the signal fingerprints collected by devices. We turned it into a logical one, and we exploited that to build a relational spatio-temporal database capable of supporting a crowdsourced collection of data. Next, we populated the database with heterogeneous cellular observations originating from multiple sources. In addition, we illustrate how the developed system allows us to properly deal with the evolution of the network configuration, e.g., by detecting cell renaming phenomena and by making it possible to correct inconsistent measurements coming from mobile devices, fostering positioning tasks. Finally, we provide a wide range of basic, spatial, and temporal analyses about the arrangement of the cellular network and its evolution over time, demonstrating how the developed system can be used to reconstruct and maintain a deep knowledge of the cellular network, possibly starting from crowdsourced information only.

A Novel Outdoor Positioning Technique Using LTE Network Fingerprints

In recent years, wireless-based fingerprint positioning has attracted increasing research attention owing to its position-related features and applications in the Internet of Things (IoT). In this paper, by leveraging long-term evolution (LTE) signals, a novel deep-learning-based fingerprint positioning approach is proposed to solve the problem of outdoor positioning. Considering the outstanding performance of deep learning in image classification, LTE signal measurements are converted into location grayscale images to form a fingerprint database. In order to deal with the instability of LTE signals, prevent the gradient dispersion problem, and increase the robustness of the proposed deep neural network (DNN), the following methods are adopted: First, cross-entropy is used as the loss function of the DNN. Second, the learning rate of the proposed DNN is dynamically adjusted. Third, this paper adopted several data enhancement techniques. To find the best positioning fingerprint and method, three types of fingerprint and five positioning models are compared. Finally, by using a deep residual network (Resnet) and transfer learning, a hierarchical structure training method is proposed. The proposed Resnet is used to train with the united fingerprint image database to obtain a positioning model called a coarse localizer. By using the prior knowledge of the pretrained Resnet, feed-forward neural network (FFNN)-based transfer learning is used to train with the united fingerprint database to obtain a better positioning model, called a fine localizer. The experimental results convincingly show that the proposed DNN can automatically learn the location features of LTE signals and achieve satisfactory positioning accuracy in outdoor environments.

Deep Learning for Fingerprint-Based Outdoor Positioning via LTE Networks

Fingerprint-based positioning techniques are a hot research topic because of their satisfactory accuracy in complex environments. In this study, we adopted the deep-learning-based long-time-evolution (LTE) signal fingerprint positioning method for outdoor environment positioning. Inspired by state-of-the-art image classification methods, a novel hybrid location gray-scale image utilizing LTE signal fingerprints is proposed in this paper. In order to deal with signal fluctuations, several data enhancement methods are adopted. A hierarchical architecture is put forward during the deep neural network (DNN) training. First, the proposed positioning technique is pre-trained by a modified Deep Residual Network (Resnet) coarse localizer which is capable of learning reliable features from a set of unstable LTE signals. Then, to alleviate the tremendous collection workload, as well as further improve the positioning accuracy, by using a multilayer perceptron (MLP), a transfer learning-based fine localizer is introduced for fine-tuning the coarse localizer. The experimental data was collected from realistic scenes to meet the requirement of actual environments. The experimental results show that the proposed system leads to a considerable positioning accuracy in a variety of outdoor environments.

A Context-Aware Model to Provide Positioning in Disaster Relief Scenarios

The effectiveness of the work performed during disaster relief efforts is highly dependent on the coordination of activities conducted by the first responders deployed in the affected area. Such coordination, in turn, depends on an appropriate management of geo-referenced information. Therefore, enabling first responders to count on positioning capabilities during these activities is vital to increase the effectiveness of the response process. The positioning methods used in this scenario must assume a lack of infrastructure-based communication and electrical energy, which usually characterizes affected areas. Although positioning systems such as the Global Positioning System (GPS) have been shown to be useful, we cannot assume that all devices deployed in the area (or most of them) will have positioning capabilities by themselves. Typically, many first responders carry devices that are not capable of performing positioning on their own, but that require such a service. In order to help increase the positioning capability of first responders in disaster-affected areas, this paper presents a context-aware positioning model that allows mobile devices to estimate their position based on information gathered from their surroundings. The performance of the proposed model was evaluated using simulations, and the obtained results show that mobile devices without positioning capabilities were able to use the model to estimate their position. Moreover, the accuracy of the positioning model has been shown to be suitable for conducting most first response activities.