Semantic Forwarding and Codebook-Enhanced Model Division Multiple Access for Satellite-Terrestrial Networks

Satellite-terrestrial communications are severely constrained by high path loss, limited spectrum resources, and time-varying channel conditions, rendering conventional bit-level transmission schemes inefficient and fragile, particularly in low signal-to-noise ratio (SNR) regimes. Semantic communication has emerged as a promising paradigm to address these challenges by prioritizing task-relevant information over exact bit recovery. In this paper, we propose a semantic forwarding-based semantic communication (SFSC) framework optimized for satellite-terrestrial networks. Specifically, we develop a vector-quantized joint semantic coding and modulation scheme, in which the semantic encoder and semantic codebook are jointly optimized to shape the constellation symbol distribution, improving channel adaptability and semantic compression efficiency. To mitigate noise accumulation and reduce on-board computational burden, we introduce a satellite semantic forwarding mechanism, enabling relay satellites to forward signals directly at the semantic level without full decoding and re-encoding. Furthermore, we design a channel-aware semantic reconstruction scheme based on feature-wise linear modulation (FiLM) to fuse the received SNR with semantic features, enhancing robustness under dynamic channel conditions. To support multi-user access, we further propose a codebook split-enhanced model division multiple access (CS-MDMA) method to improve spectral efficiency. Simulation results show that the proposed SFSC framework achieves a peak signal-to-noise ratio (PSNR) gain of approximately 7.9 dB over existing benchmarks in the low-SNR regime, demonstrating its effectiveness for robust and spectrum-efficient semantic transmission in satellite-terrestrial networks.

💡 Research Summary

The paper addresses the severe constraints of satellite‑terrestrial communications—high path loss, limited spectrum, and rapidly varying channel conditions—by introducing a semantic‑forwarding based semantic communication (SFSC) framework tailored for low‑Earth‑orbit (LEO) satellite networks. Traditional bit‑level transmission schemes aim to recover every transmitted symbol, which becomes inefficient and fragile in low‑SNR regimes typical of LEO links. Semantic communication, by contrast, focuses on delivering task‑relevant information rather than exact bits, offering a promising avenue for resource‑constrained satellite systems.

The authors propose a joint vector‑quantized (VQ) semantic coding and modulation scheme in which the semantic encoder and a learnable semantic codebook are co‑trained. The codebook maps compressed semantic vectors directly onto constellation symbols, shaping the symbol distribution to match the statistical properties of the satellite channel. This joint design improves both channel adaptability and compression efficiency, reducing the number of transmitted symbols needed for a given reconstruction quality.

To alleviate noise accumulation and the heavy computational load associated with full decode‑and‑re‑encode operations on board a satellite, the paper introduces a “semantic forwarding” mechanism. A relay satellite extracts semantic features from the received signal using a lightweight decoder, then forwards these features without performing a complete bit‑level reconstruction. This approach prevents the propagation of semantic noise across hops and cuts the on‑board model parameters by roughly 84 % compared with conventional regenerative joint source‑channel coding schemes.

At the receiver, a channel‑aware reconstruction module injects the instantaneous signal‑to‑noise ratio (SNR) into the semantic decoder via Feature‑wise Linear Modulation (FiLM). By conditioning the decoder on SNR, the system dynamically adapts its reconstruction strategy to the current channel state, markedly improving robustness under the fast‑fading, Doppler‑shifted conditions of LEO links.

For multi‑user scenarios, the authors extend the framework with a Codebook‑Split‑enhanced Model Division Multiple Access (CS‑MDMA) technique. The global semantic codebook is partitioned into a shared core and user‑specific sub‑codebooks, allowing multiple users to share model parameters while preserving distinct semantic subspaces. This design mitigates inter‑user interference without relying on strict model orthogonality or common‑information assumptions, thereby boosting spectral efficiency.

Simulation experiments employ realistic LEO channel models that incorporate multi‑path propagation, Doppler shifts, and time‑varying fading. The task is image reconstruction, evaluated by Peak Signal‑to‑Noise Ratio (PSNR) and Structural Similarity Index (SSIM). In the low‑SNR regime (≤0 dB), the SFSC framework achieves a PSNR gain of approximately 7.9 dB over state‑of‑the‑art benchmarks, while using far fewer model parameters. In multi‑user tests, CS‑MDMA outperforms traditional NOMA and RSMA schemes, delivering up to 1.5× higher spectral efficiency.

The paper also discusses limitations: the joint training requires extensive channel‑aware datasets, the optimal split ratio of the codebook is currently determined empirically, and hardware implementation aspects remain to be explored. Future work is suggested on online adaptive codebook updates, ultra‑lightweight decoder designs for on‑board deployment, and extensions to multimodal tasks such as video and speech.

In summary, the proposed SFSC framework combines vector‑quantized semantic coding, satellite‑level semantic forwarding, SNR‑conditioned reconstruction, and a novel CS‑MDMA multiple‑access scheme to deliver robust, spectrum‑efficient, and computationally lightweight semantic communication for satellite‑terrestrial networks, marking a significant step toward practical 6G non‑terrestrial integration.