VeISAC: An End-to-End MIMO-OFDM-FMCW Framework for Integrated Sensing and Communication in 6G Vehicular Networks
git clone https://github.com/mohababsa/VeISAC.git
cd VeISACconda env create -f environment.yml
conda activate veisac-envbash install_torch.shVeISAC operates on channel datasets generated using DeepVerse 6G. DeepVerse 6G produces synchronized multi-modal data (MIMO channel tensors, radar returns, LiDAR point clouds, camera frames, and vehicle trajectories) from ray-traced urban vehicular scenarios. Dataset generation scripts and scenario configurations are available on the DeepVerse 6G website under the Scenarios tab.
The dataset used in our evaluations was generated by our team using the DeepVerse 6G pipeline after configuring the Outdoor 1 (O1) scenario (https://deepverse6g.net/scenario/O1). The generated dataset is publicly available at: [Dataset link — coming soon]
Once downloaded, place the dataset under data/ and update the path in the relevant configuration file under configs/.
The simulation parameters used across all experiments are listed below.
| ISAC Shared Parameters | OFDM Communication Parameters | FMCW Radar Parameters | |||
|---|---|---|---|---|---|
| Parameter | Value | Parameter | Value | Parameter | Value |
| Frequency band | n257 (26.5–29.5 GHz) | Sampling rate F_s,c | 245.76 MHz | Sampling rate F_s,r | 200 MHz |
| Carrier frequency f_c | 28 GHz | Subcarrier spacing Δf | 120 kHz | Chirp slope μ | 2.4×10¹³ Hz/s |
| System bandwidth B | 200 MHz | FFT size N_FFT | 2048 | Samples/chirp N_samp | 1664 |
| Wavelength λ | ≈ 10.71 mm | Active subcarriers N_sc | 1633 | Number of chirps N_c | 128 |
| Transmit power P_TX | 43 dBm (20 W) | CP length N_CP | 1024 samples | Chirp duration T_chirp | 8.32 μs |
| ISAC-TX antennas M_t | 4 (2×2 UPA) | OFDM symbols/slot N_sym | 14 | RX FoV (az×el) | ±90°×±90° |
| UE Com-RX M_c | 2 (2×1 ULA) | Pilot spacing (time) D_t | 3 symbols | Sensing noise figure ℱ_s | 5 dB |
| Mono. BS Sen-RX M_r | 4 (2×2 UPA) | Pilot spacing (freq.) D_f | 3 subcarriers | Sensing noise power σ²_n,s | −86.0 dBm |
| Bist. BS Sen-RX M_r | 4 (2×2 UPA) | Total pilots/slot | 2725 | ||
| Bist. UE Sen-RX M_r | 2 (2×1 ULA) | Default modulation | QPSK | ||
| Antenna spacing d | λ/2 = 5.35 mm | RX FoV (az×el) | 360°×180° | ||
| Power allocation angle | θ = π/4 | Comm. noise figure ℱ_c | 7 dB | ||
| Power constraint | cos²θ + sin²θ = 1 | Comm. noise power σ²_n,c | −84.0 dBm | ||
| TX FoV (az×el) | 360°×180° |
Implements the full transmitter chain: QAM modulation, pilot insertion, IFFT-based OFDM generation, linear FMCW chirp synthesis, and additive time-domain superposition ith trigonometric power allocation enforcing cos²θ + sin²θ = 1.
Implements the OFDM communication receiver: CP removal, FFT demodulation, pilot-based LS/MMSE channel estimation with FMCW interference-aware regularization, MMSE equalization, power recovery, and bit demapping.
Implements the FMCW sensing receiver: NLMS-based OFDM interference mitigation, de-chirping, windowed range/Doppler FFT, zero-Doppler clutter removal, 2D OS-CFAR detection, MUSIC angle estimation, and CRB computation.
Top-level orchestrator coordinating TX, channel application, Com-RX, Sen-RX, and joint performance evaluation (PAPR, BER, throughput, range/velocity/angle errors, EIR, Equiv-MSE, Capacity-Distortion) across all three baseline topologies.
Run the full end-to-end signal processing simulation across all baselines and calculate the communication, sensing, and unified ISAC performance metrics:
conda activate veisac-env
python scripts/run_veisac.py Results and figures are saved under results/.
[1] U. Demirhan, A. Taha, S. Jiang, and A. Alkhateeb, "DeepVerse 6G: A Dataset Generation Framework for Multi-Modal Sensing and Communication Digital Twins," preprint, Feb. 2025.
[2] Remcom, "Wireless InSite," [Online]. Available: https://www.remcom.com/wireless-insite
1. VeISAC Paper: M. Ababsa, S. Ribouh, Y. El Hillali, and A. Rivenq, "VeISAC: An End-to-End MIMO-OFDM-FMCW Framework for ISAC in 6G Vehicular Networks," submitted to IEEE Transactions on Vehicular Technology, 2026.