Project description
Analysis and modeling the transmission channel is a foundation of
transmission theory with the key goal of enabling channel equalization
and optimal information decoding. Modern wireless communications, based
on OFDM modulations and (massive) MIMO technologies, achieve
unprecedented transmission speed and efficiency exploiting the
description of the channel through the CSI, enabling the receiver to
greatly improve the performance.
Transmission theory focuses on short term channel modeling, within the so-called coherence time, which is the essential descriptor that enables optimal decoding. In the past decade, however, the research community has discovered that the long term behavior of the channel, which is normally not useful for transmission performance, yields useful information on the propagation environment leading to the idea of joint communication and sensing (JCS).
Sensing includes people localization and tracking, object recognition, but also e-health and emotional computing applications. We stress that this is very different from standard device localization as nobody is asked to carry a device and the ambient is sensed as if communications signals were a radar, but without the need of ‘pulse-and-reflect’ operation.
The potential of JCS is enormous, and all future telecommunication systems, from 5G to emerging Wi-Fi standards foresee JCS in their evolution path. At the same time this potential entails also enormous privacy and security/safety concerns, as JCS is a pervasive tracking system that cannot be countered with any cryptographic means, as the information is embedded within the signal at the physical level.
CSI-Future focuses on JCS based solely on the interpretation of the CSI with AI/ML techniques, coupled with appropriate protocols to operate it and preserve safety and privacy of users.
CSI-based JCS has been tackled through AI/ML, yielding very promising results, but without completely uncovering how the information from the environment is embedded in the signal, nor proposing interpretative or predictive models for the shaping of CSI by the environment. Studies on privacy and security are still in their infancy.
The goal of CSI-Future is to push the state of the art one step further offering models that increase the insight in the problem and possible solutions to the privacy and security issues. CSI-Future starts from an extensive measurement campaign to collect a very large dataset to analyze, a feat that, strangely enough, has not yet been tackled by the community. The dataset will be published as Open Data, and fundamental stochastic analysis and modeling will be carried out. Further, the project will investigate different learning techniques to identify what are those that yield the best results in ambient sensing. Finally CSI-Future will propose protocols to exploit the results obtained in real communication systems as well as techniques to enable only authorized sensing and prevent attacks.
Partners
University of Brescia (Prof. Renato Lo Cigno) CSI-FUTURE PI
Politecnico di Torino (Prof. Carla Fabiana Chiasserini)
University of Verona (Prof. Damiano Carra)