CEDRIC

GSM trace mobile measurements are used to study indoor handset localization in an urban apartment setting. Nearest-neighbor, Support Vector Machine (SVM), and Gaussian Process classifiers are compared. A linear SVM is found to provide mean room-level classification efficiency near 100%, but only when the full set of GSM carriers is used. To our knowledge, this is the first study to use fingerprints containing all GSM carriers, and the first to suggest that GSM could be useful for very high-performance indoor localization.

A study is made of subsets of relevant GSM carriers for an indoor localization problem. A database was created containing power measurement scans of all available GSM carriers in 5 of 8 rooms of a second storey laboratory in central Paris, France, and a statistical learning algorithm developed to discriminate between rooms based on these carrier strengths. To optimize the system, carrier relevance was ranked using either Orthogonal Forward Regression or Support Vector Machine -Recursive Feature Elimination procedures, and a subset of relevant variables obtained with cross-validation. Results show that the 60 most relevant carriers are sufficient to correctly localize 97% of scans in an independent test set.

A new technique developed at ESPCI ParisTech should allow cellular received signal strength fingerprints to play an important role in localization systems for regions which are not well covered by GPS. The article describes the ARPEGEO project, initiated to evaluate the impact of full-band GSM fingerprints analyzed with modern machine learning techniques. Results on indoor localization, as well as techniques to facilitate practical implementation of the method, are presented.

Indoor handset localization in an urban apartment setting is studied using GSM trace mobile measurements. Nearest-neighbor, Support Vector Machine, Multilayer Perceptron, and Gaussian Process classifiers are compared. The linear Support Vector Machine provides mean room classification accuracy of almost 98% when all GSM carriers are used. To our knowledge, ours is the first study to use fingerprints containing all GSM carriers, as well as the first to suggest that GSM can be useful for localization of very high performance.

Currently, indoor localization is among the most challenging issues related to the Internet of Things (IoT). Most of the state-of-the-art indoor localization solutions require a high computational complexity to achieve a satisfying localization accuracy and do not meet the memory limitations of IoT devices. In this paper, we develop a localization framework that shifts the online prediction complexity to an offline preprocessing step, based on Convolutional Neural Networks (CNN). Motivated by the outstanding performance of such networks in the image classification field, the indoor localization problem is formulated as 3D radio image-based region recognition. It aims to localize a sensor node accurately by determining its location region. 3D radio images are constructed based on Received Signal Strength Indicator (RSSI) fingerprints. The simulation results justify the choice of the different parameters, optimization algorithms, and model architectures used. Considering the trade-off...

Je tiens en tout premier lieu à remercier Monsieur Bruce Denby pour avoir dirigé cette thèse et dont l'aide précieuse m'a été indispensable sur le plan scientifique. Je tiens également à le remercier pour la confiance et la sympathie qu'il m'a témoignées au cours de ces trois années de thèse. Je remercie également Monsieur Gérard Dreyfus, directeur du laboratoire SIGMA de l'ESPCI, pour m'avoir accueilli au sein de son équipe. Jʹadresse également mes remerciements à Madame Genviève Baudoin et à Monsieur Giuseppe Abreu pour avoir accepté d'être les rapporteurs de cette thèse, ainsi qu'à Messieurs Aziz Benlarbi-Delaï, Richard Dusséaux et Emmanuel Viennet de l'avoir examinée. Pendant une partie de me thèse j'ai eu la chance de travailler avec Messieurs Pierre Roussel et Rémi Dubois, Maîtres de conférences au laboratoire SIGMA, je tiens à les remercier pour leurs encouragements et soutien constant. Je remercie également tous les autres membres du laboratoire SIGMA pour leur accueil et leur bonne humeur quotidienne. Un grand merci à tous mes amis, qui se reconnaitront, de m'avoir soutenu et supporté durant ces trois années de thèse. Parce qu'elle représente beaucoup pour moi, je tiens à exprimer ma profonde gratitude à ma mère qui m'a accompagné, soutenu et encouragé tout au long de mon parcours, ainsi qu'à tous les membres de ma famille qui ont toujours cru en moi. Enfin, merci à mon époux Adlene qui m'a toujours donné le coup de pouce dont j'ai besoin dans les moments difficiles.

The article presents an indoor localization scheme for mobile devices based on GSM Received Signal Strength fingerprints combined with embedded sensor information and an area site map. Displacements of a mobile user are first estimated using a sensor dead-reckoning approach that adapts stride length to different users and environments, and a dynamically switched orientation estimation scheme responding to orientation changes of the mobile device. Positions derived from GSM fingerprints, along with constraints imposed by a site map, are then integrated using a particle filter in order to prevent the accumulation of deadreckoning errors over time. The study demonstrates that a standard handset with a cellular network connection and embedded inertial sensors can provide a good solution for indoor localization.

In this paper, we compare different order of statistical moments to blindly estimate the propagation channel in an OFDM based wireless network. We first derive the theoretical expression of the channel estimation error and we compare it to simulation results in different scenarios. These simulation results show a good agreement with the theoretical expression. Furthermore, we show that blind channel identification algorithms, based on different moment's order, exhibit similar performance in terms of the Normalized Mean Square Error (NMSE) and in terms of the Bit Error Rate (BER). We conclude that the choice of the moment's order is uniquely based on the ambiguity solving stage in the algorithm. The proposed study shows that the choice of the 4 th moment is a trade-off between channel estimation performance and ambiguity solving complexity.

This paper addresses two majors problems in today's wireless network using MIMO-OFDM. In fact, the performances of such a network depend on the interference management and on the equalization. These two problems are closely linked to channel estimation algorithms. In this paper, we propose a resource sharing scenario combined with a blind channel estimation algorithm based on fourth order statistics of the received signal. The proposed scheme is compared to the existing T/OFDM scenario and simulation results show its outperformance.

The article presents an easy to implement approach for indoor localization and navigation that combines Bayesian filtering with support vector machine classifiers to associate high-dimensionality cellular telephone network received signal strength fingerprints to distinct spatial regions. The technique employs a "space sampling" and a "time sampling" scheme in the training procedure, and the Bayesian filter allows introducing a priori information on room layout and target trajectories, resulting in robust room-level indoor localization.

In this paper, we propose a high accuracy fingerprint-based localization scheme for the Internet of Things (IoT). The proposed scheme employs mathematical concepts based on sparse representation and matrix completion theories. Specifically, the proposed indoor localization scheme is formulated as a simple optimization problem which enables efficient and reliable algorithm implementations. Many approaches, like Nesterov accelerated gradient (Nesterov), Adaptative Moment Estimation (Adam), Adadelta, Root Mean Square Propagation (RMSProp) and Adaptative gradient (Adagrad), have been implemented and compared in terms of localization accuracy and complexity. Simulation results demonstrate that Adam outperforms all other algorithms in terms of localization accuracy and computational complexity.

Location Fingerprinting (LF) is a promising localization technique that enables enormous commercial and industrial Location-Based Services (LBS). Existing approaches either appeal to the simple Received Signal Strength (RSS), which suffers from dramatic performance degradation due to sophisticated environmental dynamics, or rely on the fine-grained physical layer Channel State Information (CSI), whose intricate structure leads to an increased computational complexity. In this paper, we adopt Autoregressive (AR) modeling based entropy of CSI amplitude as location fingerprint, which shares the structural simplicity of RSS while exploiting the most location-specific statistical channel information. On this basis, we design EntLoc, a CSI-based probabilistic indoor localization system using commercial off-the-shelf Wi-Fi devices. EntLoc is deployed in an office building covering over 200 m 2. Extensive indoor scenario experiments corroborate that our proposed system yields superior localization accuracy over previous approaches even with only one signal transmitter.

With expeditious development of wireless communications, location fingerprinting (LF) has nurtured considerable indoor location based services (ILBSs) in the field of Internet of Things (IoT). For most pattern-matching based LF solutions, previous works either appeal to the simple received signal strength (RSS), which suffers from dramatic performance degradation due to sophisticated environmental dynamics, or rely on the finegrained physical layer channel state information (CSI), whose intricate structure leads to an increased computational complexity. Meanwhile, the harsh indoor environment can also breed similar radio signatures among certain predefined reference points (RPs), which may be randomly distributed in the area of interest, thus mightily tampering the location mapping accuracy. To work out these dilemmas, during the offline site survey, we first adopt autoregressive (AR) modeling entropy of CSI amplitude as location fingerprint, which shares the structural simplicity of RSS while reserving the most location-specific statistical channel information. Moreover, an additional angle of arrival (AoA) fingerprint can be accurately retrieved from CSI phase through an enhanced subspace based algorithm, which serves to further eliminate the error-prone RP candidates. In the online phase, by exploiting both CSI amplitude and phase information, a novel bivariate kernel regression scheme is proposed to precisely infer the target's location. Results from extensive indoor experiments validate the superior localization performance of our proposed system over previous approaches.

L’objectif de la these est d’ameliorer la precision de localisation des personnes (ou objets) dans les environnements ou le signal GPS est faible, voire inexistant, par exemple a l’interieur de bâtiments. L’originalite de ce travail est d’utiliser des signaux radio recus de sources exterieures pour effectuer la localisation a l’interieur des bâtiments. Durant ce travail, des mesures des puissances recues ont ete effectuees dans diverses conditions. Etant donnees les difficultes liees aux processus experimentaux (incertitude et bruit lies aux mesures), ce dernier point a constitue�� une premiere partie de ce travail de these. Le traitement et l’interpretation des mesures des puissances de signaux radio dans un environnement de propagation instable ont ete un veritable defi durant cette etude. En effet, toutes les porteuses du reseau GSM ont ete considerees et aucune hypothese n’a ete posee, a priori, sur leurs pertinences. Une seconde partie de la these a ete justement consacree a l’...

Accurately localizing users in indoor environments remains an important and challenging task. The article presents new results on room-level indoor localization, using cellular Received Signal Strength fingerprints collected with a standard cellular handset programmed to perform fast scans of the 900 and 1800 Megahertz GSM bands as a user explores an indoor environment at a normal walking pace. Support Vector Machines are used to deal with the high dimensionality of the fingerprints. The study demonstrates that an appropriately programmed standard cellular handset can provide a simple, inexpensive solution for accurate room-level indoor localization.

The article presents a simple, practical approach for indoor localization using Received Signal Strength fingerprints from the GSM network, including an analysis of the relationship between signal strength and location, and the evolution of localization performance over time. Support Vector Machine regression applied to very high dimensional fingerprints does not reveal any smooth functional relationship between fingerprints and position. Classification using Support Vector Machines however provides very good results on discriminating different rooms in an indoor environment, albeit with performance that degrades over time. Transductive inference, introduced as a means of updating models to overcome degradation over time, provides hints that accurate indoor localization can be achieved by applying classification methods to cellular Received Signal Strength fingerprints, performance robustness being maintained via model updating and refining.

A new approach to indoor localization is presented, based upon the use of Received Signal Strength (RSS) fingerprints containing data from very large numbers of cellular base stations--up to the entire GSM band of over 500 channels. Machine learning techniques are employed to extract good quality location information from these high-dimensionality input vectors. Experimental results in a domestic and an office setting are presented, in which data were accumulated over a 1-month period in order to assure time robustness. Room-level classification efficiencies approaching 100% were obtained, using Support Vector Machines in one-versus-one and one-versus-all configurations. Promising results using semi-supervised learning techniques, in which only a fraction of the training data is required to have a room label, are also presented. While indoor RSS localization using WiFi, as well as some rather mediocre results with low-carrier count GSM fingerprints, have been discussed elsewhere, th...

In the past few years, the ability of wireless network operators to monitor audience using control frames emitted by client devices has been compromised, both by legislation treating client MAC addresses as private information and by the difficulty of distinguishing genuine client frames from those arising from the Internet of Things or from certain enhanced services. Here, a deterministic model, based on characteristics of human activity and on seasonal trends, is used to reveal underlying client statistics in raw MAC-randomized WiFi Probe Request data. The method proposes a candidate conversion factor, X, between probe request counts and the client population, which offers plausible predictions on real-world datasets.

In the past few years, data privacy legislation has hampered the ability of WiFi network operators to count and map client activity for commercial and security purposes. Indeed, since client device MAC devices are now randomized at each transmission, aggregating client activity using management frames such as Probe Requests, as has been common practice in the past, becomes problematic. Recently, researchers have demonstrated that, statistically, client counts are roughly proportional to raw Probe Request counts, thus somewhat alleviating the client counting problem, even if, in most cases, ground truth measurements from alternate sensors such as cameras are necessary to establish this proportionality. Nevertheless, localizing randomized MAC clients at a network site is currently an unsolved problem. In this work, we propose a set of nine tools for extending the proportionality between client counts and Probe Requests to the mapping of client densities in real-world outdoor WiFi netw...

This paper addresses the issue of improving quality of business process (BP) models by exploiting domain knowledge. Indeed, business process models reflect the business processes of companies. The success of these processes has a direct and undeniable impact on business operations success. Managing them through their underlying models helps improving their effectiveness, consistency, and transparency. BP modeling aims at a better understanding of processes, allowing deciders to achieve strategic goals of the company. However, several studies from the literature showed that in experienced system analysts often produce low-level quality. This situation is partly due to lack of domain knowledge. In this paper, we propose to support this modeling effort with an approach that uses domain knowledge to improve the semantic quality of BP models. We suggest to use ontologies as a mean to capture domain knowledge and meta-modeling techniques to deal with BP models independently of languages in which they are expressed. Our contribution is threefold: (1) the meta-models describing both a domain ontology and a BP model are described, (2) the alignment between the concepts of both meta-models is defined and illustrated, (3) a set of Object Constraint Language mapping rules is provided. A simple case study illustrates the process.