Evaluation of gNB Monostatic Sensing for UAV Use Case

Steve Blandino, Neeraj Varshney, Jian Wang, Jack Chuang, Camillo Gentile, Nada Golmie · 2026

Read Evaluation of gNB Monostatic Sensing for UAV Use Case by Steve Blandino, Neeraj Varshney, Jian Wang, Jack Chuang, Camillo Gentile, Nada Golmie (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-02205.txt.

Introduction

Blandino and colleagues test 5G base station sensing for drones, using a monostatic setup where one base station sends and receives. Following 3GPP urban macro aerial assumptions, they build an end to end chain for multi target detection and 3D location with a 5G downlink waveform and positioning signals. Detection tops 70 percent with false alarms under 5 percent, and location errors run to a few meters. They release the 5GNRad simulator so others can reproduce the baseline.

Catalog record

CREATORSteve Blandino, Neeraj Varshney, Jian Wang, Jack Chuang, Camillo Gentile, Nada Golmie
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31815
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.02205

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Evaluation of gNB Monostatic Sensing for UAV Use Case Steve Blandino, Neeraj Varshney, Jian Wang, Jack Chuang, Camillo Gentile, Nada Golmie Abstract 3GPP Release 19 has initiated the standardization of integrated sensing and communications (ISAC), including a channel model for monostatic sensing, evaluation scenarios, and performance assessment methodologies. These common assumptions provide an important basis for ISAC evaluation, but reproducible end-to-end studies still require a transparent sensing implementation. This paper evaluates 5G New Radio (NR) base station (gNB)-based monostatic sensing for the Unmanned Aerial Vehicle (UAV) use case using a 5G NR downlink Cyclic Prefix-Orthogonal Frequency Division Multiplexing (CP-OFDM) waveform and positioning reference signals (PRS), following 3GPP Urban Macro-Aerial Vehicle (UMa-AV) scenario assumptions. We present an end-to-end processing chain for multi-target detection and 3D localization, achieving more than 70 % detection probability with less than 5 % false alarm rate, in the considered scenario. For correctly detected targets, localization errors are on the order of a few meters, with a 90th-percentile error of 4 m and 6 m in the vertical and horizontal directions, respectively. To support reproducible baseline studies and further research, we release the simulator 5GNRad , which reproduces our evaluation. I Introduction The integration of sensing into cellular networks has emerged as a key enabler for 6G, supporting use cases such as aerial surveillance, infrastructure monitoring, transportation safety, and public security [ 9 , 6 ] . Cellular deployments offer widely distributed sensing infrastructure that can complement dedicated radar systems in terms of coverage and cost, while using existing hardware and spectrum resources. Despite rapid progress in cellular sensing research [ 8 , 10 , 2 , 11 , 4 , 7 ] , performance evaluations remain fragmented. Prior work often relies on simplified propagation assumptions, such as single-bounce targets [ 8 , 10 , 2 , 11 , 4 , 7 ] and negligible clutter [ 8 , 10 , 4 , 7 ] , and on idealized hardware models that ignore antenna patterns and residual self-interference [ 8 , 10 , 2 , 11 , 4 , 7 ] . As a result, reproducible and comparable performance evaluations remain limited, and reported results are often difficult to compare objectively across studies. In parallel with the research community, 3GPP has initiated the standardization of integrated sensing and communication (ISAC) in Technical Specification Group Radio Access Network (TSG RAN) Working Group 1 (WG1), or RAN1, by defining common evaluation scenarios and assumptions. In Release-19, the ISAC channel model extends the baseline 3GPP channel model by introducing sensing targets (STs) characterized by target radar cross section (RCS), along with a dedicated clutter model [ 1 , 3 , 12 , 13 ] . However, the sensing processing chain is outside the scope of 3GPP standardization effort. Thus, while 3GPP specifies the propagation assumptions and evaluation setting used to generate the sensing channel, it does not prescribe how the received signal should be processed into detections and target estimates. In this work, we evaluate the performance of 5G NR base station (gNB) monostatic sensing for the UAV use case using an existing 5G NR downlink Cyclic Prefix-Orthogonal Frequency Division Multiplexing (CP-OFDM) waveform and the standardized downlink positioning reference signal (PRS) as the sensing signal. The evaluation is grounded in the deployment scenario and assumptions used in 3GPP Release 19 to calibrate the ISAC channel model for UAV sensing targets. On top of the standardized channel model, we develop a baseline receiver implementation that includes channel estimation, range-Doppler processing, constant false alarm rate (CFAR) detection, angle estimation, and 3D localization. We also account for practical transceiver impairments by modeling residual self-interference as an additional noise term and by considering realistic antenna patterns and array geometries. To enable reproducibility and serve as a common reference point for further research and standards studies, we release the corresponding baseline simulator, 5GNRad [ 5 ] , together with example scenarios and scripts that reproduce the full pipeline and facilitate extensions with new algorithms and assumptions. Section II presents the sensing system model, including the waveform, array architectures, and the resulting frequency-domain received-signal model. Section III then describes the 3GPP Release-19 propagation channel model used to generate the target and background responses. Section IV details the baseline processing chain that operates on the received signal to produce detections and target parameter estimates. Section V describes the simulation setup following 3GPP Urban Macro-Aerial Vehicle (UMa-AV) scenario assumptions. Section VI provides the simulation results. Section VII concludes the paper. II 5G NR Monostatic Radar Fig. 1: A TRP transmits and receives a CP-OFDM waveform, which propagates through the 3GPP channel model. We consider a single gNB operating as a monostatic full-duplex CP-OFDM radar. The transmit–receive point (TRP) denotes the physical radio site of the gNB and comprises co-located transmit and receive uniform rectangular arrays (URAs) with N antenna elements each. Sensing is performed by transmitting 5G NR positioning reference signals (PRS) and processing the received echoes from sensing targets. The end-to-end signal generation and processing flow is shown in Fig. 1 . The processing blocks in Fig. 1 are also labeled with the corresponding 5GNRad [ 5 ] function names. Let \Delta f denote the subcarrier spacing in Hz and N_{\text{sc}} the number of subcarriers in the occupied bandwidth. PRS is mapped on a comb of size K\in\{2,4,6,12\} , so only every K -th subcarrier is active. For PRS OFDM symbol index \ell and subcarrier index k , the PRS grid is s_{k,\ell}=c_{k,\ell} if k\penalty\hskip 0.80002pt\bmod\penalty\hskip 0.80002ptK=\penalty\hskip 0.80002pt0 , and s_{k,\ell}=0 otherwise. c_{k,\ell} are QPSK-modulated symbols generated from the 31st-order Gold sequence specified in Section 5.2.1 of 3GPP TS 38.211. PRS occasions are scheduled periodically with period T_{\text{PRS}} , analogous to a pulse repetition interval, and each PRS occasion contains L_{\text{PRS}} PRS symbols. A coherent processing interval (CPI) consists of N_{\text{CPI}} PRS occasions spanning time T_{\text{CPI}}=N_{\text{CPI}}T_{\text{PRS}} . The baseband OFDM waveform is obtained via Inverse Fast Fourier Transform (IFFT) and cyclic-prefix insertion per symbol. The TRP transmits a single stream with quasi-omnidirectional coverage over the forward sector. On reception, the array signals are mapped to one or more baseband streams depending on the receive architecture. For PRS resource element (k,\ell) , the transmitted array signal is: \bm{x}_{k,\ell}=\sqrt{P_{\text{TX}}}\,\bm{f}\,s_{k,\ell}\in\mathbb{C}^{N\times 1}, (1) where P_{\text{TX}} denotes the average transmit power per active PRS resource element and \bm{f}\in\mathbb{C}^{N\times 1} is the fixed transmit weight vector. After CP removal and OFDM demodulation, the N_{\text{RF}} baseband streams produced by the receive architecture are modeled as: \displaystyle\bm{y}_{k,\ell} \displaystyle=\bm{W}_{\text{RF}}^{H}\!\left(\bm{H}_{k,\ell}\bm{x}_{k,\ell}+\bm{n}_{k,\ell}+\bm{\xi}_{k,\ell}\right) \displaystyle=\sqrt{P_{\text{TX}}}\,\bm{W}_{\text{RF}}^{H}\bm{H}_{k,\ell}\bm{f}\,s_{k,\ell}+\bm{W}_{\text{RF}}^{H}\bm{n}_{k,\ell}+\bm{W}_{\text{RF}}^{H}\bm{\xi}_{k,\ell}, (2) where \bm{H}_{k,\ell}\in\mathbb{C}^{N\times N} is the effective frequency-domain monostatic MIMO channel, as defined in Section III , at subcarrier k and PRS symbol \ell , i.e., including antenna patterns and array responses, \bm{n}_{k,\ell}\sim\mathcal{CN}(\bm{0},\sigma_{n}^{2}\bm{I}) models the receiver thermal noise, and \bm{\xi}_{k,\ell}\sim\mathcal{CN}(\bm{0},\sigma_{si}^{2}\bm{I}) denotes residual self-interference (SI) leakage after cancellation. The receive architecture is described by an analog combining matrix \bm{W}_{\text{RF}}\in\mathbb{C}^{N\times N_{\text{RF}}} that maps the N element signals into N_{\text{RF}} baseband streams, as in ( 2 ). Range-Doppler processing is performed per RF chain on the streams \{y^{(i)}_{k,\ell}\}_{i=1}^{N_{\text{RF}}} , producing N_{\text{RF}} range-Doppler maps. The receiver architecture determines the number of RF chains N_{\text{RF}} . In the full-digital case, N_{\text{RF}}=N and \bm{W}_{\text{RF}}=\bm{I}_{N} , so all element signals are available in baseband and range–Doppler processing is performed per element channel. In

Read the complete text