Source-linked AI summary
Seeing Through Sparse Signals: CS-Enhanced Sensing With QoS Guarantees for 6G ISAC
Yifei Wang, Pradyumna Kumar Bishoyi, Marina Petrova
TL;DR
Sparse NR reference signals and constrained resource sharing create a challenge for weak-target detection while satisfying communication QoS. The paper compares QoS-guaranteed and PF-based scheduling and applies adaptive 2D-mmCAMP reconstruction; simulations report improved weak-target detectability while meeting DL QoS.
Problem
Sparse NR reference signals can undersample sensing channels and create sidelobes that mask weak targets, while prior scheduling work did not assess stringent communication QoS constraints.
Source: §I · p. 1Method
The paper compares QoS-guaranteed and PF-based scheduling and uses 2D-mmCAMP with adaptive row- and column-wise thresholding for sparse-reference-signal reconstruction.
Source: §I · p. 2, C. 2D Max-Median CAMP (2D-mmCAMP) · p. 4Results
At least 16% better weak-target detectability than 2D-CAMP is reported while maintaining required DL QoS and maximizing sensing performance.
Source: §I · p. 2Takeaways & Limitations
Simulations confirm that 2D-mmCAMP improves multi-target detection over the periodogram method and 2D-CAMP in the studied setting.
Source: §V · p. 6
Abstract
from arXiv · showhide
In integrated sensing and communication (ISAC) systems, communication and sensing share time-frequency resources, creating a fundamental trade-off between communication quality of service (QoS) and quality of sensing (QoSe). Reusing 5G new radio (NR) reference signals (RSs) enables standard-compliant sensing, but their inherent sparsity leads to undersampled channels and strong sidelobes that degrade weak-target detection. Compressed sensing (CS) can mitigate these effects, yet existing methods either overlook communication QoS constraints or rely on reconstruction algorithms that are step-size-sensitive or employ global thresholding, leading to weak-target masking in multi-target scenarios. The multi-target detection problem, therefore, requires a joint design of resource scheduling and CS-based reconstruction. In this work, we investigate the impact of two scheduling schemes, namely QoS-guaranteed scheduling and proportional-fairness (PF)-based scheduling, on the sensing-communication trade-off. To further enhance sensing performance, we employ a 2D max-median complex approximate message passing (2D-mmCAMP) algorithm that performs adaptive row- and column-wise thresholding to suppress sidelobes while preserving weak-target responses in multi-target scenarios. Simulations demonstrate an improvement of >= 16% in weak-target detection over baseline reconstruction algorithms while meeting downlink (DL) QoS constraints, validating the effectiveness of the proposed NR-compliant ISAC scheme.
I. INTRODUCTION
The paper jointly addresses sensing-resource scheduling under communication QoS constraints and weak-target recovery from sparse NR reference signals. It proposes QoS-guaranteed scheduling and adaptive 2D-mmCAMP reconstruction for multi-target sensing.
I. INTRODUCTION
- Sparse NR reference signals undersample sensing channels, increase sidelobes, and can mask weak targets, while prior work did not assess scheduling under stringent communication QoS constraints.The paper identifies a need to jointly design resource scheduling and reconstruction for sparse NR reference-signal patterns.Source: §I · pp. 1–2
- The proposed QoS-guaranteed scheduler maximizes target detectability while meeting a minimum downlink communication-throughput requirement, and is compared with PF-based scheduling.The study evaluates how the two scheduling schemes affect the trade-off between downlink throughput and target detection probability.Source: §I · p. 2
- 2D-mmCAMP uses adaptive row- and column-wise max-median thresholding to suppress sidelobes and improve weak-target recovery without additional computational cost.The method is intended for non-homogeneous multi-target environments with sparse NR reference-signal patterns.Source: §I · p. 2
- At least 16% better weak-target detectability than 2D-CAMP is reported while maintaining required DL QoS and maximizing sensing performance.Simulations report reliable weak-target resolution in the presence of strong targets in the Doppler domain.Source: §I · p. 2
II. SYSTEM MODEL
The system model considers a 5G base station performing monostatic sensing and downlink communication with time-frequency-multiplexed resources. It defines target detection QoSe and uses detection probability as the sensing KPI.
II. SYSTEM MODEL
- The model uses a 26 GHz 5G base station for simultaneous monostatic sensing and downlink communication, with sensing and communication scheduled in a time-frequency-multiplexed manner.It assumes full-duplex operation with sufficiently low self-interference to be cancelled.Source: §II · p. 2
- Sensing and communication resource elements partition the available grid, with their allocated counts summing to the total number of resource elements.The total grid has Nsc × Nsym elements.Source: §II · p. 2
A. Target Detection QoSe
- Target detection QoSe is deciding whether a target is present from noisy observations and includes resolving and identifying multiple targets.The base station processes echoes reflected from the target.Source: A. Target Detection QoSe · p. 2
- The model uses detection probability as its sensing KPI, with decisions based on received SNR and a specified false-alarm probability.The received echo power and thermal noise determine the sensing SNR; the detection-probability expression uses the first-order Marcum Q-function.Source: A. Target Detection QoSe · p. 2
B. Communication QoS
Downlink throughput is the model's communication KPI, calculated over the communication resource elements during the observation period.
B. Communication QoS
- Downlink throughput is calculated using the allocated communication resource elements over the observation period, with symbol duration including the cyclic prefix.The throughput expression is given in bits/s.Source: B. Communication QoS · p. 2
C. Sensing Signal Model
The base station schedules NR reference signals for sensing, estimates the channel from received echoes, and generates a range-Doppler map for target detection. The signal model relates target delay and Doppler to distance and radial velocity.
C. Sensing Signal Model
- The BS schedules NR reference signals according to sensing and communication requirements, then uses the resulting channel estimate to generate a range-Doppler map for detection.The BS is assumed to perform monostatic sensing and downlink communication simultaneously.Source: C. Sensing Signal Model · p. 3
- The received echoes represent K point targets, with each target characterized by complex channel gain, round-trip delay, and Doppler shift.The delay and Doppler shift correspond to target distance and radial velocity, respectively.Source: C. Sensing Signal Model · p. 3
- NR reference signals create a sparse binary pattern in the time-frequency grid, so channel estimates are retained only at sensing-allocated resource elements.The estimate is set to zero at other resource elements.Source: C. Sensing Signal Model · p. 3
III. COMPRESSED SENSING-AUGMENTED SCHEDULING
The scheduling analysis represents resource contention between sensing and communication through their shares of the available resource elements. The sensing fraction is β, while the remaining fraction is distributed among downlink users according to their QoS requirements.
III. COMPRESSED SENSING-AUGMENTED SCHEDULING
- β denotes the fraction of total resource elements allocated to sensing, leaving 1 − β for communication.The communication share can be distributed among multiple downlink users according to their QoS requirements.Source: §III · p. 3
- The analysis assumes that sensing and communication compete only within the resource-element domain.Source: §III · p. 3
A. QoS-Guaranteed Scheduling
QoS-guaranteed scheduling maximizes target detection probability subject to a minimum downlink throughput requirement. Because detection probability increases monotonically with the sensing resource share, the optimum uses the largest share compatible with that requirement.
A. QoS-Guaranteed Scheduling
- The optimal sensing allocation is β = 1 − (Creq · Nsym · T)/(NRE · log2(1 + γcom)), which maximizes detection probability while meeting the downlink throughput requirement.The detection probability is stated to increase monotonically with β.Source: A. QoS-Guaranteed Scheduling · p. 3
B. PF-based Scheduling
PF-based scheduling maximizes the sum of the logarithms of sensing and communication performance, with the logarithm preventing either objective from overwhelming the other. The section also motivates CS reconstruction for sparse sensing allocations and introduces adaptive noise-floor estimation to address weak-target masking.
B. PF-based Scheduling
- The PF objective maximizes the sum of the logarithms of sensing and communication performance, yielding a unique optimal sensing allocation because the optimization is convex.The logarithm prevents one objective from overwhelming the other.Source: B. PF-based Scheduling · p. 4
- With heavy downlink requirements, low sensing-resource allocation produces decimated channel estimates and reduced sensing performance.Source: B. PF-based Scheduling · p. 4
- 2D-mmCAMP uses adaptive noise-floor estimation for each range-Doppler bin to mitigate weak-target masking by nearby strong-target sidelobes.The paper motivates this adaptation because a common median estimate can miss weak targets in non-homogeneous multi-target scenarios.Source: B. PF-based Scheduling · p. 4
C. 2D Max-Median CAMP (2D-mmCAMP)
2D-mmCAMP adapts its denoising threshold to each delay-Doppler cell using row- and column-wise medians, while retaining 2D-CAMP’s computational complexity.
C. 2D Max-Median CAMP (2D-mmCAMP)
- 2D-mmCAMP uses adaptive row- and column-wise thresholding for multi-target scenarios without increasing total computational complexity over 2D-CAMP.Both algorithms have complexity O(niterNscNsym log(NscNsym)).Source: C. 2D Max-Median CAMP (2D-mmCAMP) · p. 4
- For each delay-Doppler cell, 2D-mmCAMP thresholds using the larger of its row and column medians, unlike 2D-CAMP’s common median across all elements.It then applies element-wise soft-thresholding and uses the residual in the next iteration.Source: C. 2D Max-Median CAMP (2D-mmCAMP) · p. 4
IV. PERFORMANCE EVALUATION
The evaluation compares communication throughput and detection probability across scheduling and sensing conditions, using a modeled mmWave propagation environment.
IV. PERFORMANCE EVALUATION
- Figure 3 presents the tradeoff between DL communication throughput and probability of detection for QoS-guaranteed and PF-based scheduling.The two panels vary communication SNR and required throughput, respectively.Source: Figure 3 · p. 5
- Figure 4 plots detection probability against velocity spacing between targets.Source: Figure 4 · p. 5
- The simulations use a line-of-sight channel and MATLAB ray tracing to model mmWave propagation.The scenario is presented in Fig. 2.Source: §IV · p. 5
A. Communication and Sensing Tradeoff Evaluation
The evaluation examines how scheduling affects the throughput–detection tradeoff and tests CS reconstruction for weak-target detection under sparse NR reference-signal conditions. 2D-mmCAMP improves detection most when targets are closely spaced in Doppler.
A. Communication and Sensing Tradeoff Evaluation
- For γcom ≤ 2 dB, QoS-guaranteed scheduling achieves higher DL throughput than PF-based scheduling at the cost of lower sensing performance.At γcom = −2 dB, βopt ≈ 1/6 yields PD,opt = 0.15; PF scheduling has PD,PF = 0.58 but its throughput falls below Creq at low communication SNR.Source: A. Communication and Sensing Tradeoff… · p. 5
- With a good communication channel and fixed sensing SNR, PD,opt is around 0.98 at relatively low Creq and falls to 0.88 at Creq = 100 Mbps, while PD,PF = 0.69 for all Creq.Source: A. Communication and Sensing Tradeoff… · p. 5
B. Multi-target Detectability with CS
- The multi-target comparison tests 2D-mmCAMP against periodogram and 2D-CAMP using two same-distance targets with different radial velocities and RCSs.The weak target has lower RCS; the strong target has larger RCS.Source: B. Multi-target Detectability with CS · p. 5
1) Effect of Doppler Separation:
- The Doppler-spacing test fixes weak- and strong-target RCS at 5 and 20 dBsm, respectively, and varies their velocity difference to assess masking.Source: 1) Effect of Doppler Separation · p. 5
- At vdiff = 2.5, 2D-mmCAMP improves weak-target detection by approximately 10% over 2D-CAMP and 2D-FFT for PRS6, and up to 15% for DMRS3.When vdiff > 3, masking becomes negligible and the three methods’ performance converges.Source: 1) Effect of Doppler Separation · p. 5
2) Low-RCS Target Detectability:
For closely spaced targets, 2D-mmCAMP improves weak-target detection over 2D-CAMP and 2D-FFT. The conclusion places this sensing improvement within NR-compliant resource sharing under QoS-guaranteed and PF-based scheduling.
2) Low-RCS Target Detectability:
- For weak targets below 0 dBsm, 2D-mmCAMP increases PD by up to 34% over 2D-CAMP and 38% over 2D-FFT.For PRS6, it raises PD by up to 15% over 2D-CAMP; for DMRS3, PD reaches 1 at 5 dBsm, versus 0.85 for 2D-FFT and 2D-CAMP.Source: 2) Low-RCS Target Detectability · p. 6
- For both PRS6 and DMRS3, 2D-mmCAMP produces a more distinct weak-target peak than 2D-FFT or 2D-CAMP in the two-target heatmaps.The comparison uses targets at 86 m, with RCSs of 20 and 7 dBsm and velocity separation vdiff = 2.5; adaptive row- and column-wise thresholding improves detection probability.Source: 2) Low-RCS Target Detectability · p. 6, Figure 6 · p. 6
V. CONCLUSIONS
- The study compares QoS-guaranteed and PF-based scheduling; the former has a closed-form optimal RE allocation, while the latter's unique optimum is obtained numerically.Under limited sensing resources, the conclusion reports using 2D-mmCAMP to reconstruct high-resolution range-Doppler images and mitigate weak-target masking from strong-target sidelobes.Source: §V · p. 6
- Simulations confirm that 2D-mmCAMP improves multi-target detection over the periodogram method and 2D-CAMP; future work will incorporate multipath models and evaluate different schedulers.Source: §V · p. 6