PaperlayerView exact revision on arXiv

Source-linked AI summary

Toward Reliable and Accurate Predictive ISAC in Mobile mmWave Networks

Atefeh Rezaei, Vahid Jamali, Falko Dressler

arXiv:2609.29741v1eess.SP

TL;DR

CSI training overhead is costly in mmWave systems, motivating predictive ISAC that constructs partial CSI from sensed LoS and NLoS components. The paper combines this channel approach with worst-user-CRB beamforming and a non-coherent design; it reports maintaining connectivity when LoS links are unavailable by exploiting backscattered echoes.

  • Problem

    Large antenna arrays and sparse mmWave propagation make conventional CSI estimation and 5G-NR approaches costly in pilot overhead.

    Source: §I · p. 1

  • Method

    The framework uses predicted sensing parameters to construct partial LoS/NLoS CSI and designs worst-user-CRB beamforming, including a robust variant without absolute phase knowledge.

    Source: §I · p. 2

  • Results

    The framework reports maintaining connectivity when LoS links are unavailable by using blockage detection and backscattered echoes.

    Source: §VI · p. 14

  • Takeaways & Limitations

    The framework combines geometry-based channel modeling and EKF-enabled prediction for mobile mmWave tracking and localization with LoS and mobile-scatterer NLoS reflections.

    Source: §VI · p. 14

Abstract

from arXiv · show

Integrated sensing and communications (ISAC) systems offer a promising framework for beam tracking, which enhances channel awareness and improves communication reliability in mobile millimeter wave (mmWave) networks. Motivated by this, we propose a low-complexity, model-driven channel state information (CSI) framework with integrated channel prediction capability in ISAC-enabled systems. To reduce the overhead of CSI acquisition and uplink feedback, the proposed framework leverages predicted target sensing parameters to construct partial CSI from both line-of-sight (LoS) and detected non-line-of-sight (NLoS) components. Exploiting this partial CSI, we formulate a beamforming optimization problem that minimizes the worst targets' Cramér-Rao bound (CRB) under per-user communication constraints. This problem is non-convex and highly non-linear; therefore, we develop an efficient suboptimal solution using successive convex approximation (SCA) and Schur-complement-based decomposition techniques, and we evaluate its complexity in comparison with other approaches. Moreover, since obtaining the absolute phase of the channel paths may require additional pilot resources, we formulate a robust beamforming design that does not assume this knowledge, and propose a solution based on sample averaging. Numerical results demonstrate that the ISAC-based robust channel prediction and localization framework significantly improves prediction accuracy and link reliability compared with conventional schemes, such as communication-based localization and tracking and separated sensing and communication. These results highlight the potential of ISAC for reliable, adaptive, and high-frequency, blockage-prone wireless networks. Furthermore, the performance of the proposed framework is evaluated under severe blockage scenarios, which further demonstrates enhanced network reliability.

I. INTRODUCTION

In mobile mmWave networks, costly CSI training motivates predictive ISAC that uses sensed geometry to recover channel information from LoS and detected NLoS paths. The paper develops robust beamforming and compares its channel model and optimization with prior predictive tracking approaches.

I. INTRODUCTION

  • Predicted sensing parameters construct partial CSI from LoS and detected NLoS components, supporting geometry-aware channel prediction and robust next-slot beamforming.The model also includes mobile-scatterer reflections and uses an EKF for CSI prediction.Source: §I · pp. 1–2
  • The design minimizes the worst-user CRB under per-user rate constraints using an epigraph- and Schur-complement reformulation with SCA.The resulting non-convex min-max formulation directly designs dual-functional beamforming vectors rather than power allocation or time splitting.Source: §I · p. 2, §II · p. 3
  • The robust non-coherent beamforming design avoids assuming absolute channel-path phase, which may require extra pilots, signaling, and feedback.It addresses phase rotations associated with carrier frequency offset, multipath fading, Doppler, and time-synchronization errors.Source: §I · p. 2
  • The evaluation examines performance under mobile, non-coherent conditions and highlights the role of NLoS links when LoS paths are blocked.The paper also evaluates gains from robust optimization.Source: §I · p. 2

II. RELATED WORK

  • Prior beam-tracking studies predominantly assume LoS channels, while several optimize data rates with limited attention to whether reflected signals support reliable sensing.The stated gap motivates channel models and optimization that account for NLoS components and sensing quality.Source: §II · p. 3
  • Unlike 's Sub-6 GHz fallback after blockage, the framework uses detected mobile scatterers to reconstruct NLoS components and continue serving blocked users over mmWave.The paper distinguishes its combination of NLoS-aware CSI modeling, non-coherent beamforming, blockage detection, and interference-aware prediction from the compared approaches.Source: §II · p. 3

III. SYSTEM MODEL

The DFRC base station uses a single waveform to communicate with mobile users and sense reflected signals. Its radar receive beam is steered using predicted vehicle angles to improve target reception and suppress clutter.

III. SYSTEM MODEL

  • The DFRC-BS serves mobile users while sensing vehicles that reflect its signals, and those vehicles can also act as scatterers for nearby users.The system has separate transmit and receive ULAs, with Nt transmit antennas and Nr receiving elements.Source: §III · p. 3

A. Radar Model

  • The radar receive beam is steered toward the predicted vehicle direction to attenuate clutter from other directions.MUSIC can further distinguish user signals from clutter and estimate multiple arrival angles.Source: §III · p. 3, A. Radar Model · p. 4
  • Each time slot carries a user-specific signal with a dual-functional transmit beam, and the BS processes target echoes using a receive beamformer.The radar model assumes the target’s round-trip channel remains fixed throughout a slot.Source: A. Radar Model · p. 4

B. Communication Receiver Model

The DFRC signal carries user data while the system processes echoes to estimate and predict channel parameters. This closes a per-slot loop that produces beamforming for the next slot.

B. Communication Receiver Model

  • A user’s received signal includes a direct path and reflections, and its channel model includes LoS and NLoS components estimated from tracked-user parameters.The achievable rate is expressed as Rk,n = log(1 + Γk,n).Source: B. Communication Receiver Model · p. 4

IV. PROPOSED PREDICTIVE ISAC FRAMEWORK

  • Each slot uses a DFRC block for simultaneous data transmission and echo processing, then predicts channel state and beamforming parameters for the next slot.The estimates are obtained using matched filtering and MUSIC, then refined by an EKF before beamforming is optimized.Source: §IV · p. 5

A. Tracking Module

The tracking module estimates vehicle angles, delays, and Doppler shifts, then uses an EKF to predict their next-slot states. The channel model uses predicted geometry to reconstruct LoS and detected NLoS components, avoiding exhaustive user feedback.

A. Tracking Module

  • MUSIC estimates target angles, matched filtering estimates delay and Doppler, and an EKF tracks and predicts vehicle angle, distance, and velocity.Vehicles are assumed synchronized to the BS clock.Source: A. Tracking Module · p. 5

B. Channel Modeling Module

  • The channel model reconstructs detectable LoS and NLoS paths from tracked-target geometry, while residual undetected scatterers remain stochastic.The framework predicts the next-slot channel from predicted geometry and avoids user feedback for exhaustive multipath estimation.Source: B. Channel Modeling Module · pp. 5–6
  • The predicted channel treats dominant paths as quasi-static within a slot after Doppler compensation, while residual unknown or undetected scatterers are modeled stochastically.Predictions for slot n+1 are based on observations and estimates from slot n.Source: B. Channel Modeling Module · p. 7

C. Optimization Module

The optimization minimizes the worst target’s CRB while meeting user rate and power constraints. Epigraph reformulation, Schur complements, and SCA yield a convex iterative subproblem.

C. Optimization Module

  • The design minimizes the worst-angle CRB while enforcing a minimum user data rate and a shared power budget for communication and sensing.The original min-max problem is computationally complex because its objective and rate constraint are non-convex.Source: C. Optimization Module · p. 7, 1) Solution Methodology · p. 7

1) Solution Methodology:

  • Epigraph reformulation and a Schur-complement constraint transformation help address the non-convex min-max problem.The rate constraint is approximated by a convex constraint using a Taylor expansion, while SCA handles remaining non-convexity.Source: 1) Solution Methodology · pp. 7–8
  • The resulting P3 is convex and jointly optimizes beamforming vectors, slack variables, and an epigraph variable through iterative SCA updates solved with CVX.The algorithm then estimates echoes with matched filtering and MUSIC, refines target parameters with an EKF, and reconstructs the next channel.Source: 1) Solution Methodology · p. 8
  • Detected scatterers are used to improve network performance under blockages.The framework includes blockage detection using received backscattered echoes.Source: 1) Solution Methodology · p. 8

2) Robust non-coherent optimization:

The robust design treats absolute path phase as uniformly distributed uncertainty and uses sample averaging to enforce communication-rate constraints across phase realizations.

2) Robust non-coherent optimization:

  • Absolute path phase is modeled as uniformly distributed on [0, 2π) to account for residual uncertainty from synchronization, feedback, hardware, and propagation factors.The passage links mmWave motion of a few millimeters to significant phase shifts and rapid absolute-phase fluctuations.Source: 2) Robust non-coherent optimization · pp. 8–9
  • The design requires achievable data rates to hold for all possible values of the random phase parameter.Source: 2) Robust non-coherent optimization · p. 9
  • Sample average approximation draws U independent phase samples, converting the stochastic rate constraint into a deterministic convex problem using an SCA-based approximation.Source: 2) Robust non-coherent optimization · p. 9
  • Increasing the sample count U improves approximation accuracy and captures phase uncertainty more fully in the constraint.Source: 2) Robust non-coherent optimization · p. 9

D. Overall Algorithm

Algorithm 1 presents the alternating-optimization-based solution.

D. Overall Algorithm

  • Algorithm 1 presents the AO-based solution; subsequent sections analyze its convergence and computational complexity.Source: D. Overall Algorithm · p. 9

1) Convergence Analysis:

The SCA iterations preserve feasibility, decrease the objective sequence, and converge to a stationary point of the original problem.

1) Convergence Analysis:

  • First-order Taylor expansions form convex surrogates at each SCA iteration, with convergence argued from feasibility, monotonicity, and boundedness.Source: 1) Convergence Analysis · p. 9
  • Inner approximations keep each iterate feasible for the original problem, while minimizing the objective makes objective values non-increasing and convergent to a finite value.Source: 1) Convergence Analysis · p. 9
  • The algorithm converges to a stationary point; the auxiliary variable and main beamforming variable are optimized simultaneously to satisfy constraints and preserve SCA convergence guarantees.Source: 1) Convergence Analysis · p. 9

2) Computational Complexity:

The complexity analysis specifies problem-size and constraint counts for P3 and contrasts the SCA formulation with SDR-based lifting.

2) Computational Complexity:

P SDR

The paper reports optimization complexity per time slot and evaluates the framework in a roadway scenario that includes denser traffic and modeled environmental scatterers.

P SDR

  • For P5, single-slot complexity is reported as PS = KNt + K + 1 and CS = k(2 + U) + 1; frame complexity multiplies this by the number of time slots.The passage describes the proposed SCA approach as reducing computational cost, particularly with large antenna arrays.Source: P SDR · p. 9

V. SIMULATION RESULTS

  • The evaluation covers detection and tracking accuracy, achievable data rate, blockage detection, and robust beamforming under absolute phase ambiguity.Source: §V · p. 9

A. Simulation Scenario

  • The roadway setup grows from K = 3 baseline vehicles to K = 9, with additional scatterers modeled by Rayleigh fading at power-splitting coefficient ˜γ = 0.8.The baseline vehicles travel at 35–55 km/h; denser traffic includes six additional vehicles traveling at 60–85 km/h.Source: A. Simulation Scenario · pp. 9–10
  • Scalability tests extend to K = 35 users, while a separate scenario increases vehicle velocities by up to 30 km/h during tracking.The K = 35 users are randomly distributed along the road segment.Source: A. Simulation Scenario · pp. 9–10

B. Estimation accuracy and tracking validation

The proposed DFRC approach outperforms communication-based tracking in angle and range, while benchmark comparisons show tradeoffs against sensing-only and interference-unaware designs. Accuracy worsens with denser, faster traffic, and higher sensing power improves angle-estimation RCRB.

B. Estimation accuracy and tracking validation

C. Communication performance evaluation

The evaluation treats spectral efficiency as secondary to worst-user sensing accuracy and examines communication behavior under changing SINR, blockage, and phase uncertainty. It reports benefits from modeled NLoS links and robust non-coherent beamforming.

C. Communication performance evaluation

VI. CONCLUSION

  • The framework combines geometry-based LoS/NLoS channel modeling and EKF prediction with worst-CRB beamforming and blockage detection for mobile mmWave V2I.Source: §VI · pp. 13–14

D. Blockage detection effect

E. Robust design

  • Under random phase, robust non-coherent beamforming yields higher achievable rates than non-robust allocation, while coherent estimation with ideal phase knowledge performs better still.Source: E. Robust design · p. 13, Figure 13 · p. 14
Loading 2609.29741v1…