Semantic communication based on generative AI: a new approach to image compression and edge optimization

As digital technologies continue to advance, modern communication networks face unprecedented challenges in handling the vast amounts of data produced daily by connected intelligent devices. Autonomous vehicles, smart sensors, IoT systems etc., are gaining more and more interest and new communication paradigms are needed. This thesis addresses these challenges by combining semantic communication with generative models to optimize image compression and resource allocation in edge networks. Unlike traditional bit-centric communication systems, semantic communication prioritizes the transmission of meaningful data specifically selected to convey the meaning rather than obtain a faithful representation of the original data. The communication infrastructure can benefit of the focus solely on the relevant parts of the data due to significant improvements in bandwidth efficiency and latency reduction. Central to this work is the design of semantic-preserving image compression algorithms, utilizing advanced generative models such as Generative Adversarial Networks and Denoising Diffusion Probabilistic Models. These algorithms compress images by encoding only semantically relevant features and exploiting the generative power at the receiver side. This allows for the accurate reconstruction of high-quality images with minimal data transmission. The thesis also introduces a Goal-Oriented edge network optimization framework based on the Information Bottleneck problem and stochastic optimization, ensuring that communication resources are dynamically allocated to maximize efficiency and task performance. By integrating semantic communication into edge networks, the proposed system achieves a balance between computational efficiency and communication effectiveness, making it particularly suited for real-time applications. The thesis compares the performance of these semantic communication models with conventional image compression techniques, using both classical and semantic-aware evaluation metrics. The 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.