Semantic Communication based on Generative AI: A New Approach to Image Compression and Edge Optimization

TL;DR

This work advances semantic communication by integrating generative AI with edge-network optimization to compress images semantically rather than bit-for-bit. It introduces modular SPIC and SQ-GAN frameworks that couple semantic maps with coarse reconstructions and utilize diffusion and VQ-based generative models to preserve meaning at low bitrates. A GoAL resource-allocation scheme based on Information Bottleneck and stochastic optimization demonstrates how to adapt compression and computation at the edge under latency and performance constraints. Extensive evaluation on semantic metrics (e.g., mIoU, ACC) and classic metrics on Cityscapes-like data shows superior semantic preservation at lower bitrates compared to traditional codecs, highlighting practical impact for autonomous systems and real-time edge inference. The work also provides a unified framework combining IB-based task relevance with diffusion/GAN-based reconstruction, enabling adaptive, real-time optimization in data-driven, resource-constrained networks.

Abstract

As digital technologies advance, communication networks face challenges in handling the vast data generated by intelligent devices. Autonomous vehicles, smart sensors, and IoT systems necessitate new paradigms. This thesis addresses these challenges by integrating semantic communication and generative models for optimized image compression and edge network resource allocation. Unlike bit-centric systems, semantic communication prioritizes transmitting meaningful data specifically selected to convey the meaning rather than obtain a faithful representation of the original data. The communication infrastructure can benefit to significant improvements in bandwidth efficiency and latency reduction. Central to this work is the design of semantic-preserving image compression using Generative Adversarial Networks and Denoising Diffusion Probabilistic Models. These models compress images by encoding only semantically relevant features, allowing for high-quality reconstruction with minimal transmission. Additionally, a Goal-Oriented edge network optimization framework is introduced, leveraging the Information Bottleneck principle and stochastic optimization to dynamically allocate resources and enhance efficiency. By integrating semantic communication into edge networks, this approach balances computational efficiency and communication effectiveness, making it suitable for real-time applications. The thesis compares semantic-aware models with conventional image compression techniques using classical and semantic evaluation metrics. Results demonstrate the potential of combining generative AI and semantic communication to create more efficient semantic-goal-oriented communication networks that meet the demands of modern data-driven applications.

Figures (38)

• Figure 1: Diagram illustrating the three levels of communication and their interconnection as proposed in WARREN1953semantic
• Figure 2: Semantic Communication scheme.
• Figure 3: Impact of varying channel conditions on semantic reconstruction and how the generative model adapt to these changes. As the channel quality improves, the generative model is able to receive more semantic symbols $\mathbf{z}_i$ and eventually reach semantically equivalence between $\mathbf{x}$ and $\hat{\mathbf{x}}$.
• Figure 4: Goal-Oriented Communication Scheme.
• Figure 5: Trade-off representation of the ib problem in the Relevance-Complexity plane Zaidi2020IB. The blue line represents the optimal solutions for different values of $\beta$ while the red line represents the limit of relevance given by all the information that $\mathbf{x}$ has on $\mathbf{y}$. The optimal solution is achieved in the case of gib where the projection matrix $\mathbf{A}$ is selected as in Eq. \ref{['eq: SEMCOM Matrice_A']}.