Energy-Efficient Probabilistic Semantic Communication Over Space-Air-Ground Integrated Networks

Zhaohui Yang, Zhouxiang Zhao, Zhaohui Yang 0001, Mingzhe Chen, Chen Zhu, Wei Xu, Zhaoyang Zhang, Kaibin HuangIEEE Trans. Wireless Communications 2025 · DOI

TL;DR

This paper tackles the challenge of optimizing energy efficiency in future wireless networks that combine satellites, aerial vehicles (UAVs), and ground devices (SAGINs). They introduce a novel Probabilistic Semantic Communication (PSC) framework that compresses data by focusing on its meaning, significantly reducing communication energy at the cost of computation. The authors develop an iterative algorithm to jointly optimize communication and computation resources for maximum energy savings.

What Problem Does This Paper Address?

The rapid evolution of wireless networks is leading towards integrated Space-Air-Ground Integrated Networks (SAGINs), offering global coverage and advanced services. However, efficiently managing communication and computation resources within these complex, multi-layered networks remains a significant hurdle. Specifically, while semantic communication promises to reduce data transmission needs by conveying meaning rather than raw bits, it introduces substantial computational overhead. This paper addresses the critical need to balance these competing demands for energy efficiency in SAGINs, where communication resources are particularly valuable.

Methods

The core of this work is the development of a Probabilistic Semantic Communication (PSC) framework for SAGINs. Here's a breakdown of their approach:

  • Probabilistic Semantic Communication (PSC): Instead of transmitting raw data, PSC extracts "semantic triples" (head entity, relation, tail entity) and further compresses them using shared probability graphs. This probabilistic approach allows for omitting relation information, which the receiver can reconstruct based on the shared knowledge, leading to significant data size reduction.
  • Joint Communication and Computation Optimization: The problem is formulated as a complex optimization problem aiming to minimize total energy consumption (communication + computation). This involves optimizing several parameters simultaneously:
  • Semantic Compression Ratio: How much to compress the data semantically.
  • Computation Task Allocation: Deciding whether the satellite or UAV performs the semantic compression.
  • Computation Capacity Allocation: How much processing power to assign to each task.
  • Power and Bandwidth Allocation: Distributing these resources.
  • UAV Altitude and Location: Optimizing the UAV's position for better coverage and reduced latency.
  • Iterative Algorithm: To tackle the non-convex and non-smooth nature of the optimization problem, the authors propose an iterative algorithm. This algorithm breaks down the problem into six subproblems, solving each iteratively until convergence. Key to this is obtaining closed-form solutions for computation capacity allocation and UAV altitude at each step.
  • Relaxation and Dual Methods: For certain discrete optimization subproblems (like task allocation), they employ relaxation techniques (e.g., treating binary variables as continuous) and then use dual decomposition methods to find optimal solutions.

Datasets

Not discussed in the paper. The authors use simulated data based on system parameters.

Key Insights and Ideas

  • The Trade-off is Key: The most crucial insight is the inherent trade-off between communication and computation in semantic communication. Achieving higher compression (saving communication energy) directly leads to higher computational demands. The paper's success hinges on finding the optimal balance point.
  • Probabilistic Compression is Powerful: By leveraging shared probability graphs, the system can go beyond simple semantic extraction and probabilistically infer missing information, leading to even greater compression and energy savings.
  • Holistic Optimization is Necessary: Optimizing only communication or computation in isolation is insufficient. The paper demonstrates the effectiveness of jointly optimizing task allocation, compression ratios, resource allocation, and UAV positioning for maximum energy efficiency.
  • UAV Location Matters, But Less Than Expected: While optimizing UAV location is part of the problem, the results suggest its impact on overall energy consumption is less significant than other factors like satellite beam gain or data size, especially when compared to the benefits of semantic compression.

Key Results

  • The proposed SAGIN-PSC scheme consistently achieves the lowest total energy consumption compared to non-semantic, random computation allocation, and fixed UAV location schemes, across various data sizes and satellite beam gains.
  • The algorithm converges rapidly, typically within three iterations, demonstrating its efficiency.
  • The PSC technique significantly reduces energy consumption, with savings outweighing the additional computational energy cost, especially when an optimal semantic compression ratio is applied.
  • The paper provides closed-form solutions for computation capacity allocation and UAV altitude, simplifying the optimization process.

Limitations and Future Work

  • Scenario Expansion: The authors suggest expanding the scenario to include multiple satellites and UAVs, which would introduce further complexity and require more sophisticated coordination.
  • Dynamic Elements: Future work could incorporate dynamic elements such as satellite motion and more complex UAV trajectories to enhance robustness and adaptability.
  • Real-world Deployment: While simulations show promise, real-world implementation would need to account for more complex channel models, interference, and practical hardware constraints.

Summary updated

BibTeX
@article{yang2025energy,
  author = {Zhaohui Yang and Zhouxiang Zhao and Zhaohui Yang and Mingzhe Chen and Chen Zhu and Wei Xu and Zhaoyang Zhang and Kaibin Huang},
  title = {{Energy-Efficient Probabilistic Semantic Communication Over Space-Air-Ground Integrated Networks}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2025},
  doi = {10.1109/twc.2025.3569102},
}
Figure