Clutter-Aware Integrated Sensing and Communication: Models, Methods, and Future Directions

This repository contains the MATLAB simulation code for the paper:

R. Liu, P. Li, M. Li, and A. L. Swindlehurst, "Clutter-aware integrated sensing and communication: Models, methods, and future directions," Proc. IEEE, to appear, 2026. [arXiv]

Overview

We develop a unified wideband MIMO-OFDM signal model that captures both cold clutter (environmental backscatter of the probing waveform) and hot clutter (external interference reflections) across the space, time, and frequency domains. The code implements the complete simulation framework including:

• Clutter channel generation: cold clutter with frequency-correlated coefficients, hot clutter via bistatic scattering paths, and UAV-like extended targets (Section IV)
• Slow-time clutter suppression: symbol-wise averaging (Avg.), recursive mean averaging (RMA), and consecutive symbol differencing (CSD) (Section V-A)
• Spatial-domain suppression: MVDR beamforming under ideal, insufficient-training, and AoA-mismatch scenarios (Section V-B)
• Space-time adaptive processing (STAP): conventional, full-rank, and reduced-rank STAP for joint angle-Doppler filtering (Section V-C)
• Clutter-aware transceiver co-design: BLP-based alternating optimization with Dinkelbach-SDP for joint transmit covariance and receive beamformer design under communication QoS constraints (Section VI)
• Clutter kernel estimation: probing-based estimation of the spatial clutter kernel and its projection onto the completely positive (CP) cone (Section IV)

Requirements

• MATLAB R2022a or later (tested on R2023a)
• CVX (version 2.2 or later) with a compatible SDP solver - http://cvxr.com/cvx/
• MOSEK (recommended, used in the Dinkelbach-SDP solver) - https://www.mosek.com/
  • If MOSEK is unavailable, change cvx_solver mosek to cvx_solver sdpt3 or cvx_solver sedumi in function/solve_TX_Dinkelbach.m
• Parallel Computing Toolbox (optional but recommended)
  • Several scripts use parfor for Monte Carlo acceleration. If the toolbox is unavailable, replace parfor with for - the code will run correctly but more slowly.

Repository Structure

code/
├── README.md                       # This file
├── param_basic.m                   # System parameter initialization (run first)
├── Vcc_Reta_esti.m                 # Clutter kernel and covariance estimation
├── BLP_BFstap_Design.m             # BLP-based joint transceiver design
│
├── fig2_RDM_derandom.m             # Fig. 2:  Range-Doppler maps (de-randomization)
├── fig4_5_slow_time_Processing.m   # Fig. 4-5: Slow-time suppression + MUSIC spectra
├── fig8_spatial_MVDR.m             # Fig. 8:  Spatial MVDR beampatterns
├── fig10_STAP_sim.m                # Fig. 10: Angle-Doppler maps (STAP)
├── fig12a_SCNR_Ptx.m               # Fig. 12a: SCNR vs. transmit power
├── fig12b_SCNR_Gamma.m             # Fig. 12b: SCNR vs. communication SINR threshold
│
├── function/                       # Supporting functions
│   ├── gen_clutter_channels.m      # Generate cold/hot/UAV clutter channel tensors
│   ├── draw_H_cold_realization.m   # Draw one random cold clutter realization
│   ├── echo_tar.m                  # Generate target echo signal
│   ├── clutter_reflectivity_GIT.m  # GIT-type empirical clutter reflectivity model
│   ├── scatterAngles.m             # Generate scatterer angles with target avoidance
│   │
│   ├── symbol_wise_mean_esti.m     # Slow-time: symbol-wise mean subtraction (Avg.)
│   ├── rma_filter.m                # Slow-time: recursive mean averaging filter (RMA)
│   ├── frame_wise_difference.m     # Slow-time: consecutive symbol differencing (CSD)
│   ├── music_spectrum.m            # MUSIC spatial spectrum computation
│   ├── mvdr_beamformer.m           # MVDR receive beamformer with diagonal loading
│   │
│   ├── kernel_apply.m              # Apply clutter kernel: Rcc = Vcc * vec(Rxx)
│   ├── build_Ac.m                  # Build clutter-dependent matrix for TX optimization
│   ├── project_Vcc_to_CP.m         # Project clutter kernel onto the CP cone (via Choi matrix)
│   ├── compute_SCNR.m              # Compute overall SCNR across subcarriers
│   ├── comp_SINR_comm.m            # Compute per-user communication SINR
│   │
│   ├── solve_BLP_AO.m              # ISAC: AO-based joint TX/RX design (BLP + MVDR)
│   ├── solve_Radar_AO.m            # Radar-only: AO-based TX/RX design
│   ├── solve_TX_Dinkelbach.m       # Dinkelbach-SDP solver (ISAC with QoS constraints)
│   ├── solve_TX_Dinkelbach_radar.m # Dinkelbach-SDP solver (radar-only)
│   ├── baseline_comm_heur.m        # Comm-only and heuristic ISAC baselines
│   ├── recover_beams.m             # Recover beamformers from covariance solutions
│   │
│   ├── annotRectData.m             # Annotation: rectangle in data coordinates
│   ├── annotEllipseData.m          # Annotation: ellipse in data coordinates
│   └── localData2Norm.m            # Convert data coordinates to normalized figure units
│
└── data/                           # Generated data  
    ├── parameters_basic.mat        # Output of param_basic.m
    └── covMat_BLP.mat              # Output of Vcc_Reta_esti.m

Quick Start

Step 1: Generate basic parameters

run('param_basic.m')

This creates data/parameters_basic.mat containing the MIMO-OFDM system configuration, target parameters, hot source settings, and clutter region geometry.

Step 2: Generate individual figures

The following scripts can be run independently after Step 1:

Script	Paper Figure	Description
fig2_RDM_derandom.m	Fig. 2	Range-Doppler maps with and without de-randomization
fig4_5_slow_time_Processing.m	Fig. 4 & 5	MUSIC spectra and RDMs under slow-time suppression
fig8_spatial_MVDR.m	Fig. 8	MVDR beampatterns (ideal, insufficient training, AoA mismatch)
fig10_STAP_sim.m	Fig. 10	Angle-Doppler maps (conventional, full-rank STAP, reduced-rank STAP)

Step 3: Run BLP-based transceiver design (requires CVX)

% Step 3a: Estimate clutter kernel and covariance matrices (compute-intensive)
run('Vcc_Reta_esti.m')

% Step 3b: Run the joint transceiver design
run('BLP_BFstap_Design.m')

Step 4: Reproduce SCNR performance curves (requires CVX, compute-intensive)

% Requires covMat_BLP.mat from Step 3a
run('fig12a_SCNR_Ptx.m')     % Fig. 12a: SCNR vs. transmit power
run('fig12b_SCNR_Gamma.m')   % Fig. 12b: SCNR vs. communication SINR threshold

Note: Vcc_Reta_esti.m, fig12a_SCNR_Ptx.m, and fig12b_SCNR_Gamma.m involve large-scale Monte Carlo simulations and may take several hours to complete. Using parfor with multiple workers is highly recommended.

System Parameters

The default parameters in param_basic.m correspond to a 28 GHz mmWave MIMO-OFDM ISAC system:

Parameter	Value	Description
Carrier frequency	28 GHz	mmWave band
Subcarrier spacing	120 kHz	5G NR numerology μ = 3
Number of subcarriers	512	Total bandwidth ≈ 61.44 MHz
Number of OFDM symbols	256	Coherent processing interval
Transmit/Receive antennas	16 / 16	ULA with half-wavelength spacing
Transmit power	53 dBm
Communication users	3
Target RCS	-13 dBsm	Point target
Clutter scatterers	100	Distributed in two range rings

Citation

If you use this code in your research, please cite:

@article{liu2026clutter,
  author  = {Liu, Rang and Li, Peishi and Li, Ming and Swindlehurst, A. Lee},
  title   = {Clutter-Aware Integrated Sensing and Communication: Models, Methods, and Future Directions},
  journal = {Proc. IEEE},
  year    = {2026},
  note    = {to appear}
}

Contact

• Rang Liu - University of California, Irvine - rangl2@uci.edu
• Peishi Li - Dalian University of Technology - lipeishi@mail.dlut.edu.cn

License

This code is provided for academic and research purposes. Please contact the authors for commercial use.