Andrea Giorgetti

Blind Selection of Representative Observations for Sensor Radar Networks

Sensor radar networks enable important new applications based on accurate localization. They rely on the quality of range measurements, which serve as observations for inferring a target location. In harsh propagation environments (e.g., indoors), such observations can be nonrepresentative of the target due to noise, multipath, clutter, and non-line-of-sight conditions leading to target misdetection, false-alarm events, and inaccurate localization. These conditions can be mitigated by selecting and processing a subset of representative observations. We introduce blind techniques for the selection of representative observations gathered by sensor radars operating in harsh environments. A methodology for the design and analysis of sensor radar networks is developed, taking into account the aforementioned impairments and observation selection. Results are obtained for noncoherent ultra-wideband sensor radars in a typical indoor environment (with obstructions, multipath, and clutter) to enable a clear understanding of how observation selection improves the localization accuracy.

Combined Finite-Discrete Numerical Modeling of Runout of the Torgiovannetto di Assisi Rockslide in Central Italy

AbstractThe combined finite–discrete-element method (FDEM) is an advanced and relatively new numerical modeling technique that combines the features of the FEM with those of the discrete-element method. It simulates the transition of brittle geomaterials from continua to discontinua through fracture growth, coalescence, and propagation. With FDEM, it is possible to simulate landslides from triggering to runout and carry out landslide scenario analyses, the results of which can be successively adopted for cost-effective early warning systems. The purpose of this paper is to describe the results of the FDEM simulations of the triggering mechanism and the evolution scenarios of the Torgiovannetto di Assisi rockslide (central Italy), a depleted limestone quarry face where a rock wedge with an approximate volume of 182,000 m3 lies in limit equilibrium conditions, posing relevant issues in terms of civil protection. The results obtained demonstrate that the FDEM is able to realistically simulate the different p...

A Robust Wireless Sensor Network for Landslide Risk Analysis: System Design, Deployment, and Field Testing

In this paper, we propose a wireless sensor network (WSN) designed for monitoring and risk management of landslides, where data collected by sensors are delivered through the network to a remote unit for online analysis and alerting. To ensure fast deployment, robustness in harsh environments, and very long lifetime, the sensor nodes and the communication protocol have been specifically conceived, so that the network is self-organizing, fault tolerant, and adaptive. The WSN has been installed on a landslide located in Torgiovannetto (Italy) for an experimental campaign of several months where performance metrics, such as radio link and path statistics as well as battery levels, have been collected. These metrics demonstrated the effectiveness of the network protocols to manage self-organization, node failures, low link quality, and unexpected battery depletion. With negligible human intervention during the pilot experiment, the WSN revealed a very high level of robustness, which makes it suitable to monitor landslides in critical scenarios.

Constrained cluster based blind localization of primary user for cognitive radio networks

Blind localization of primary user (PU) is a geo-location spectrum awareness feature that can be very useful in enhancing the functionality of cognitive radios (CRs) in terms of minimizing the interference to the PU. However, the estimation of the PU position within the region is made difficult because cooperation between the PU and the secondary user (SU) does not exist and therefore the PU signal parameters remain unknown to the SU. The centroid-based localization techniques have significantly been adopted as suitable candidates that do not require knowledge of such parameters. In this paper we investigate the localization performance of such techniques by imposing constraints to the selection of the SU nodes, termed as SU cluster, to estimate the PU location. In particular, we impose a minimum distance constraint between any two SU nodes and group the qualifying nodes into a cluster. Only the SU nodes from the constrained cluster can take part in localizing the PU. We simulate the proposed method for a shadow fading wireless environment and compare the results with the centroid and the weighted centroid based blind localization methods. Our results show that the mean squared error in the estimation of the position of the PU is significantly improved for the proposed method compared to the two standard centroid localization techniques especially when the true PU location is away from the center of the region.

Analysis of the Restricted Isometry Property for Gaussian Random Matrices

In the context of compressed sensing, we provide a new approach to the analysis of the symmetric and asymmetric restricted isometry property for Gaussian measurement matrices. The proposed method relies on the exact distribution of the extreme eigenvalues for Wishart matrices, or on its approximation based on the Tracy-Widom law, which in turn can be approximated by means of properly shifted and scaled Gamma distributions. The resulting probability that the measurement submatrix is ill conditioned is compared with the known concentration of measure inequality bound, which has been originally adopted to prove that Gaussian matrices satisfy the restricted isometry property with overwhelming probability. The new analytical approach gives an accurate prediction of such probability, tighter than the concentration of measure bound by many orders of magnitude. Thus, the proposed method leads to an improved estimation of the minimum number of measurements required for perfect signal recovery.

On sparse recovery using finite Gaussian matrices: Rip-based analysis

We provide a probabilistic framework for the analysis of the restricted isometry constants (RICs) of finite dimensional Gaussian measurement matrices. The proposed method relies on the exact distribution of the extreme eigenvalues of Wishart matrices, or on its approximation based on the gamma distribution. In particular, we derive tight lower bounds on the cumulative distribution functions (CDFs) of the RICs. The presented framework provides the tightest lower bound on the maximum sparsity order, based on sufficient recovery conditions on the RICs, which allows signal reconstruction with a given target probability via different recovery algorithms.

Indoor detection and tracking of human targets with UWB radar sensor networks

In this paper, the whole signal processing chain for an ultra-wideband (UWB) radar sensor network (RSN) is presented, starting from real measurements collected by sensor devices, and ending with the estimation of the target trajectory. The RSN is composed by one transmitter and six receivers, and monitors an indoor area of about 70 m2 performing tracking of a human target. The proposed processing chain consists of: clutter removal, a novel detection scheme, one-dimensional clustering, trilateration, two-dimensional clustering, particle-based probability hypothesis density filter tracking, and data association. Numerical results show the remarkable performance of the system, resulting in a root mean square localization error of 36 cm, a value smaller than the target size. The presented experimental study show that it is possible to accurately track human targets using a UWB RSN in a densely cluttered environment.

Exact analysis of weighted centroid localization

Source localization of primary users (PUs) is a geolocation spectrum awareness feature that can be very useful in enhancing the functionality of cognitive radios (CRs). When the cooperating CRs have limited information about the PU, weighted centroid localization (WCL) based on received signal strength (RSS) measurements represents an attractive low-complexity solution. In this paper, we propose a new analytical framework to calculate the exact performance of WCL in the presence of shadowing, based on results of the ratio of two quadratic forms in normal variables. In particular, we derive an exact expression for the root mean square error (RMSE) of the two-dimensional location estimate. Numerical results confirm that the derived framework is able to predict the performance of WCL capturing all the essential aspects of propagation as well as CR network spatial topology.

Statistical distribution of position error in weighted centroid localization

Weighted centroid localization (WCL) based on received signal strength (RSS) measurements is an attractive low-complexity solution that enables cognitive radios (CRs) to have a geolocation awareness of the radio environment. In this paper, we propose a new analytical framework to accurately calculate the performance of WCL based on the statistical distribution of the ratio of two quadratic forms in normal variables. In particular, we derive an exact expression for the cumulative distribution function (CDF) of the two-dimensional location estimation in the presence of independent and identically distributed (i.i.d.) as well as correlated shadowing. Numerical results confirm that the analytical framework is able to predict the performance of WCL capturing all the essential aspects of propagation as well as CR network spatial topology.

Weak RIC Analysis of Finite Gaussian Matrices for Joint Sparse Recovery

This letter provides tight upper bounds on the weak restricted isometry constant for compressed sensing with finite Gaussian measurement matrices. The bounds are used to develop a unified framework for the guaranteed recovery assessment of jointly sparse matrices from multiple measurement vectors. The analysis is based on the exact distribution of the extreme singular values of Gaussian matrices. Several joint sparse reconstruction algorithms are analytically compared in terms of the maximum support cardinality ensuring signal recovery, i.e., mixed norm minimization, MUSIC, and OSMP based algorithms.