CRC-Assisted Channel Codes for Integrated Passive Sensing and Communications

TL;DR

This work tackles coded passive ISAC by introducing CIPSAC, a framework that uses CRC-assisted coding with OFDM to enable reliable communications while simultaneously guiding passive sensing at a multi-antenna BS. The core methodology, IPSCD, iteratively refines both sensing and decoding by treating CRC-passed data as extra pilots and by replacing wrongly decoded packets with zeros to preserve sensing integrity; the approach leverages an SOS of techniques including a theoretical optimality result for the zero-codeword replacement and K-best decoding for NOS codes. A learning-based NOS coding scheme is introduced, trained with a loss that balances sensing accuracy and data reliability, enabling superior performance in short block-length regimes. The framework integrates a detailed sensing pipeline (AoA, delay, Doppler, and RCS) with a joint channel estimation and decoding loop, showing notable gains in both mean-squared sensing error and packet error rate through a few iterations, supported by ablation studies. Overall, the work demonstrates that coding-aware passive ISAC can achieve substantial gains by jointly optimizing codes, iterative detection, and sensing algorithms, with practical implications for 6G-like systems employing OFDM and multi-antenna receivers.

Abstract

We propose a novel coded integrated passive sensing and communication (CIPSAC) system with orthogonal frequency division multiplexing (OFDM), where a multi-antenna base station (BS) passively senses the parameters of the targets and decodes the information bit sequences transmitted by a user. The transmitted signal is comprised of pilot and data OFDM symbols where the data symbols adopt cyclic redundancy check (CRC)-assisted channel codes to facilitate both the decoding and sensing procedures. In the proposed scheme, CRC not only enhances the reliability of communication but also provides guidance to the parameter sensing procedure at the BS. In particular, a novel iterative parameter sensing and channel decoding (IPSCD) algorithm is proposed, where the correctly decoded codewords that pass CRC are utilized for sensing to improve the parameter estimation accuracy, and in return, more accurate parameter estimates lead to a larger number of correctly decoded data symbols. Conventional sensing algorithms rely only on the received pilot signals, while we utilize both the data and pilot signals for sensing. We provide a detailed analysis of the optimal strategy, in which the wrongly decoded data packets are replaced by zero codewords. To further improve the performance, we introduce learning-based near-orthogonal superposition (NOS) codes, which exhibit superior error correction capability especially in the short block length regime. NOS codes are trained using a weighted loss function, where a hyper parameter is introduced to balance the sensing and the communication losses. Simulation results show the effectiveness of the proposed CIPSAC system and the IPSCD algorithm, where both the sensing and decoding performances are significantly improved with a few iterations. We also carry out extensive ablation studies for a comprehensive understanding of the proposed scheme.

Figures (3)

• Figure 1: In the considered CIPSAC system, the user transmits both pilot and coded data packets to the BS using OFDM. We assume $L$ mobile targets/scatterers, where the $\ell$-th one, $\ell \in [1, L]$, is characterized by its radar cross section (RCS), $\alpha_\ell$, delay, $\tau_\ell$, Doppler, $\mu_\ell$, and AoA, $\theta_\ell$, parameters.
• Figure 2: Illustration of the proposed CIPSAC system. At the encoder, $M_d$ bit sequences are first CRC encoded followed by channel encoding and modulation to generate the data symbols. At the decoder, we run the proposed IPSCD algorithm where for the $n$-th iteration, the channel decoder produces both the decoded bit sequence, $\hat{\bm{c}}_b$ and the CRC flag, $T_b$. The decoded bit sequence $\bm{\hat{c}}_b$ with $T_b = 1$ are re-encoded yet the failed ones with $T_b = 0$ are replaced by randomly generated signals for the next round of parameter sensing process.
• Figure 3: Illustration of the sensing procedure of the proposed IPSCD algorithm. In particular, we first apply MUSIC algorithm to the input signal ${\bm{Y}}$ to obtain the AoA estimates, $\hat{\theta}_\ell, \ell \in [1, L]$. Then, for the $\ell$-th AoA estimate, $\hat{\theta}_\ell$, we perform spatial filtering, data substitution, and data equalization detailed in Section \ref{['sec:sense_alg']} to obtain $\bm{Z}_\ell \in \mathbb{C}^{N \times (M_p + M_d)}$. After applying the 2D-FFT to $\bm{Z}_\ell$, the delay and Doppler parameters, $\hat{\tau}_\ell, \hat{\mu}_\ell$ (or equivalently, $\hat{n}_\ell, \hat{m}_\ell$), associated with the AoA estimate, $\hat{\theta}_\ell$ can be obtained. Finally, the RCS parameters are calculated according to Eqn. \ref{['equ:est_alpha']}.

Theorems & Definitions (4)

• Proposition 1
• proof
• Remark 1
• Remark 2