A Joint JSCC-Resource Allocation Framework for QoS-Aware Semantic Communication in LEO Satellite-based EO Missions

TL;DR

This paper presents a joint optimization framework combining semantic communication and resource allocation for LEO satellite Earth observation missions, significantly reducing power consumption while maintaining image quality.

Contribution

It introduces a novel joint JSCC-resource allocation method with a curve-fitting model to optimize power efficiency under QoS constraints in satellite communications.

Findings

Achieves substantial power savings over traditional methods.

Effectively balances image quality and transmission power.

Demonstrates robustness in dynamic LEO satellite conditions.

Abstract

In Earth observation (EO) missions with Low Earth orbit (LEO) satellites, high-resolution image acquisition generates a massive data volume that poses a significant challenge for transmission under the limited satellite power budget, while LEO movement introduces dynamic systems. To enable efficient image transmission, this paper employs semantic communication (SemCom) with joint source-channel coding (JSCC), which focuses on transmitting meaningful information to reduce power consumption. Under a quality-of-service (QoS) requirement defined by image reconstruction quality, this work aims to minimize the total transmit power by jointly optimizing the JSCC encoder-decoder parameters and resource allocation. However, the implicit relationship among JSCC parameters, link quality, and image quality, coupled with the presence of mixed integer-continuous variables, makes the problem difficult to solve directly. To address this, a curve-fitting model is proposed to approximate the JSCC compression-SNR-quality relationship. Then, the joint compression ratio-resource allocation (JCRRA) algorithm is proposed to address the underlying problem. Numerical results demonstrate that the proposed method achieves substantial power savings compared to both greedy algorithms and conventional transmission paradigms.