In this work, we propose a new framework for blind wireless network topology inference and present a novel solution based on machine learning (ML) techniques. In particular, we seek to identify a causal relationship between the patterns of the radio-frequency (RF) transmissions of the nodes in the network from over-the-air signals observed by a cloud of sensors randomly deployed in the network landscape. The proposed framework is based on simple RF sensors that measure the received power at a rate sufficient to extract traffic patterns. Numerical results based on simulated data show how, despite the propagation impairments and noise may affect the performance of the algorithms, the neural network (NN)-based solution reaches 93% of accuracy even with a relatively low number of sensors.

This study explores the promising potential of integrating sensing capabilities into multiple-input multiple-output (MIMO)-orthogonal frequency division multiplexing (OFDM)-based networks through innovative multi-sensor fusion techniques, tracking algorithms, and resource management. A novel data fusion technique is proposed within the MIMO-OFDM system, which promotes cooperative sensing among monostatic (JSC) base stations by sharing range-angle maps with a central fusion center. To manage data sharing and control network overhead introduced by cooperation, an excision filter is introduced at each base station. After data fusion, the framework employs a three-step clustering procedure combined with a tracking algorithm to effectively handle point-like and extended targets. Delving into the sensing/communication trade-off, resources such as transmit power, frequency, and time are varied, providing valuable insights into their impact on the overall system performance. Additionally, a sophisticated channel model is proposed, accounting for complex urban propagation scenarios and addressing multipath effects and multiple reflection points for extended targets like vehicles. Evaluation metrics, including (OSPA), downlink sum rate, and bit rate, offer a comprehensive assessment of the system's localization and communication capabilities, as well as network overhead.

Detection of changes in indoor areas and controlled environments is getting increasing interest in ambient intelligence and security. In this paper, we propose a radio-frequency (RF)based anomaly detector that, observing the spectrum received from signals of opportunity (SoOp) and exploiting machine learning (ML) techniques, is capable of revealing changes in an indoor environment. Based on real waveforms emitted by a WiFi access point (AP) and collected by a RF sensor, we demonstrate that anomaly detection, e.g., represented by the presence of a person in the monitored area, is possible. The proposed methodology, tested in a typical office environment when the AP-sensor link is in non-line-of-sight (NLOS), achieves an accuracy greater than 95% just by collecting few beacon packets, i.e., in dozens of milliseconds. Moreover, results demonstrate that the proposed approach outperforms a well-known received signal strength (RSS)-based solution in terms of accuracy, even using just a single sensor.

The massive and autonomous structural health monitoring (SHM) of bridges is a problem that is of growing interest due to its importance and topicality. However, a considerable amount of data must be elaborated and managed in such an application. This paper proposes a set of machine learning (ML) tools to detect anomalies in a bridge from vibrational measurements using the minimum amount of data. The proposed framework starts from the fundamental frequencies extracted through operational modal analysis (OMA) and clustering, followed by a density-based time-domain tracking algorithm. The fundamental frequencies extracted are then fed to one-class classification (OCC) algorithms that perform anomaly detection. Then, to reduce the amount of data, we analyze the effect of the number of sensors, the number of bits per sample, the observation time, and the measurement noise on damage detection performance. As a case study, the Z-24 bridge is considered because of the extensive database of accelerometric measurements in both standard and damaged conditions. A comparison of OCC algorithms, such as principal component analysis (PCA), kernel principal component analysis (KPCA), Gaussian mixture model (GMM) and one-class classifier neural network (OCCNN) $$^2$$2 is performed, and their robustness to data shrinking is evaluated. In many cases, OCCNN $$^2$$2 increases the performance with respect to classical anomaly detection techniques in terms of accuracy.