Situational Awareness in Mobile Networks : Advances in Environment Sensing and Near-field RIS-aided Localization

Abstract

As mobile networks evolve from the fifth generation (5G) toward the sixth generation (6G), wireless systems are transitioning from communication-oriented infrastructures toward platforms capable of integrated sensing and communications (ISAC). This dissertation investigates key enabling technologies for situational awareness in mobile networks, addressing fundamental challenges spanning conventional far-field localization, communication-based sensing, and emerging near-field localization paradigms enabled by reconfigurable intelligent surfaces (RISs).

The first part of the dissertation examines the feasibility and performance limits of three-dimensional localization and sensing within 5G New Radio (NR) networks under far-field propagation assumptions. In particular, the work studies, describes, and evaluates 3D positioning capabilities by considering and comparing vector antennas (VAs) and uniform rectangular arrays (URAs) under different estimation frameworks. To overcome the inherent resolution limitations of compact VAs, the motion of the observer is exploited to synthesize a virtual array, thereby significantly enhancing angular estimation accuracy. Building on this localization framework, the dissertation further investigates environment sensing capabilities by processing reflections of 5G NR reference signals to form synthetic aperture radar (SAR)-like scatterer maps of ground objects in an airport surveillance use case. This enables the reconstruction of scatterer maps using standardized 5G reference signals. Unlike classical radar waveforms, the considered reference signals exhibit a non-uniform time domain allocation pattern, which leads to the mirror images of scatterers – an aspect that can be remedied via the VA based processing approach.

While the above far-field and communication-signal-based methods establish a practical baseline for localization and sensing, the evolution of 6G networks introduces new challenges that fundamentally alter the underlying propagation regime. Therefore, the second part of the dissertation shifts focus to RIS-aided localization in 6G networks, where increasing carrier frequencies and large RIS apertures give rise to radiating near-field propagation characterized by spherical wavefronts and spatial non-stationarity (SNS). In such scenarios, conventional planar-wavefront models become inaccurate. To address this challenge, the dissertation introduces a deliberate model-misspecification framework, in which the complex SNS channel is approximated by a simplified model that enables low-complexity localization algorithms.

Localization performance is evaluated in terms of root mean square error (RMSE) results, and is complemented by a rigorous misspecification-aware analytical framework based on the so-called Misspecified Cramér–Rao Bound (MCRB), which quantifies the impact of model mismatch and provides a theoretical benchmark under the model misspecification. Moreover, when a coarse prior estimate of the UE position is available, the dissertation optimizes the RIS codebook based on this coarse location, further improving localization accuracy and robustness. In particular, the proposed RIS codebook achieves sub-centimeter localization accuracy under low transmit power conditions, compared to sub-meter accuracy with unoptimized codebooks.