The proliferation of generative AI services, from cloud-based synthesis to on-device content creation, is increasingly bottlenecked by the need for efficient communication over dynamic wireless channels. However, deploying these models at the network edge is impeded by two fundamental challenges, namely the weak coupling between the model and physical channel conditions and the prohibitive computational latency of diffusion-based inference. Existing approaches often treat diffusion as a static module trained at a fixed noise level and indexed by abstract timesteps without physical interpretation, which yields suboptimal performance under time-varying channels. To address this, we propose an signal-to-noise ratio guided latent diffusion model (SNR-LDM). The diffusion process is reparameterized with a physically interpretable mapping between denoising timesteps and channel SNR, aligning the denoising trajectory with the channel state and enabling SNR-adaptive inference. To enhance semantic fidelity, we introduce multimodal guidance that fuses textual and visual prompts to steer content generation. We further develop two inference schemes, namely a dynamic multi-step procedure for highest reconstruction quality and a single-step analytic inversion for ultra-low latency, which makes deployment on resource-constrained edge devices feasible. Experiments show that the proposed model achieves about 4 dB higher PSNR than adaptive baselines and yields up to 8.2% lower LPIPS than diffusion-based benchmarks. The framework links large-scale generative modeling with the communication-constrained network edge and enables robust and efficient deployment of generative AI services.

Efficient video transmission is essential for seamless communication and collaboration within the visually-driven digital landscape. To achieve low latency and high-quality video transmission over a bandwidth-constrained noisy wireless channel, we propose a stable diffusion (SD)-based goal-oriented semantic communication (GSC) framework. In this framework, we first design a semantic encoder that effectively identify the keyframes from video and extract the relevant semantic information (SI) to reduce the transmission data size. We then develop a semantic decoder to reconstruct the keyframes from the received SI and further generate the full video from the reconstructed keyframes using frame interpolation to ensure high-quality reconstruction. Recognizing the impact of wireless channel noise on SI transmission, we also propose an SD-based denoiser for GSC (SD-GSC) condition on an instantaneous channel gain to remove the channel noise from the received noisy SI under a known channel. For scenarios with an unknown channel, we further propose a parallel SD denoiser for GSC (PSD-GSC) to jointly learn the distribution of channel gains and denoise the received SI. It is shown that, with the known channel, our proposed SD-GSC outperforms state-of-the-art ADJSCC, Latent-Diff DNSC, DeepWiVe and DVST, improving Peak Signal-to-Noise Ratio (PSNR) by 69%, 58%, 33% and 38%, reducing mean squared error (MSE) by 52%, 50%, 41% and 45%, and reducing Fréchet Video Distance (FVD) by 38%, 32%, 22% and 24%, respectively. With the unknown channel, our PSD-GSC achieves a 17% improvement in PSNR, a 29% reduction in MSE, and a 19% reduction in FVD compared to MMSE equalizer-enhanced SD-GSC. These significant performance improvements demonstrate the robustness and superiority of our proposed methods in enhancing video transmission quality and efficiency under various channel conditions.

The emergence of the Metaverse brings new opportunities for enhancing productivity and creativity through real time updates and personalized content. However, it also leads to the generation of massive volumes of dynamic information, placing unprecedented demands on existing communication networks. Current bit-oriented communication systems are not designed to cope with such high levels of semantic complexity and data volume, ultimately limiting the responsiveness and interactivity of Metaverse applications. To address this research gap, we propose a goal-oriented semantic communication framework enabled by generative AI and optimal transport for Metaverse construction (GSC). The proposed GSC framework includes an hourglass network-based (HgNet) encoder to extract semantic information of objects in the Metaverse and a semantic decoder to construct the Metaverse content after wireless transmission,enabling efficient communication and real-time object behaviour updates to the scenery for the Metaverse construction task. To overcome the wireless channel noise at the receiver, we design an optimal transport (OT)-enabled semantic denoiser, which enhances the accuracy of the Metaverse scenery through wireless communication. The results of our computer experiments demonstrate that compared to the conventional Metaverse construction, our proposed GSC framework significantly reduces wireless Metaverse construction latency by 92.6%, while improving the Metaverse object status accuracy and viewing experience by45.6% and 44.7%, respectively.

The rapid growth of multimodal data in 6G-enabled applications demands a paradigm shift from traditional bit-rate-oriented communication to semantic transmission. While diffusion models offer a promising solution for high-fidelity generation, their application is hindered by a fundamental disconnect from the physical channel and high computational overhead. Existing methods typically treat the denoising process as independent of the channel conditions, using abstract integer timesteps that lack physical meaning, and require a full, lengthy sampling chain for reconstruction. To address these challenges, we propose signal-to-noise ratio-guided latent diffusion model (SNR-LDM), which re-parameterizes the diffusion process to map denoising time steps directly to physical SNR. This enables adaptive inference from noise levels precisely matched to the estimated channel state, significantly reducing redundant computations while further guiding the reconstruction process through multimodal prompts to ensure semantic consistency. Experimental results demonstrate that, across a wide range of channel conditions, our proposed SNR-LDM achieves a PSNR gain of approximately 4 dB over ADJSCC while simultaneously reducing the LPIPS from 0.75 to 0.25.

Unmanned aerial vehicle (UAV) wireless image transmission faces challenges of limited bandwidth and dynamic channel conditions. To address these issues, we propose a novel semantic communication architecture employing a dual-level semantic transmission strategy that decomposes original images into coarse-grained and fine-grained semantic information. Specifically, by integrating the efficient MobileViT network into a joint source-channel coding (JSCC) framework, we design a lightweight semantic encoder-decoder dedicated to coding the image details. To ensure reliable transmission, we introduce a decoupled control-data transmission (DCDT) mechanism that transmits the dual-level semantic information over independent channels, with a fusion module at the receiver integrating them to produce the reconstructed image. Experimental results demonstrate that the proposed system significantly outperforms the traditional method and standard JSCC method in both reconstruction quality and channel robustness, while achieving performance comparable to the heavyweight model with substantially reduced model complexity, thereby validating its deployment potential on resource-constrained UAVs.

Efficient image transmission is essential for seamless communication and collaboration within the visually-driven digital landscape. To achieve low latency and high-quality image reconstruction over a bandwidth-constrained noisy wireless channel, we propose a stable diffusion (SD)-based goal-oriented semantic communication (GSC) framework. In this framework, we design a semantic autoencoder that effectively extracts semantic information (SI) from images to reduce the transmission data size while ensuring high-quality reconstruction. Recognizing the impact of wireless channel noise on SI transmission, we propose an SD-based denoiser for GSC (SD-GSC) conditional on an instantaneous channel gain to remove the channel noise from the received noisy SI under known channel. For scenarios with unknown channel, we further propose a parallel SD denoiser for GSC (PSD-GSC) to jointly learn the distribution of channel gains and denoise the received SI. It is shown that, with the known channel, our SD-GSC outperforms state-of-the-art ADJSCC and Latent-Diff DNSC, improving Peak Signal-to-Noise Ratio (PSNR) by 32 % and 21 %, and reducing Fréchet Inception Distance (FID) by 40 % and 35 %, respectively. With the unknown channel, our PSD-GSC improves PSNR by 8 % and reduces FID by 17 % compared to MMSE equalizer-enhanced SD-GSC.

Efficient video transmission is a critical task within the realm of current wireless communication system. Addressing the imperative to alleviate the transmission burden and conserve communication resources, we propose an optical flow-based goal-oriented semantic communication framework for video transmission (OF-GSC). Our framework features an optical flow-based semantic encoder that includes a motion extractor for optical flow estimation and a patch-level optical flow-based semantic information (SI) extractor to effectively identify and select important SI, thereby reducing the transmission load while ensuring the high-quality video reconstruction in semantic decoder. Specifically, the first original frame is leveraged as the base knowledge. Once the base knowledge and the received SI are embedded at the receiver, the embedded data is then fed into the customized autoencoder model within the semantic decoder of OF-GSC framework for efficient video reconstruction. In comparison to DeepJSCC, our OF-GSC framework achieves a significant improvement in generated video quality, as evidenced by a 13.47% increase in the Structural Similarity Index Measure (SSIM) score. Under stringent communication constraints, OFGSC surpasses M-JPEG by 14.04% in SSIM score. These results highlight the robustness and superiority of our proposed OF-GSC in efficient video transmission.

With the emergence of the metaverse and its role in enabling real-time simulation and analysis of real-world counterparts, an increasing number of personalized metaverse scenarios are being created to influence entertainment experiences and social behaviors. However, compared to traditional image and video entertainment applications, the exact transmission of the vast amount of metaverse-associated information significantly challenges the capacity of existing bit-oriented communication networks. Moreover, the current metaverse also witnesses a growing goal shift for transmitting the meaning behind custom-designed content, such as user-designed buildings and avatars, rather than exact copies of physical objects. To meet this growing goal shift and bandwidth challenge, this paper proposes a goal-oriented semantic communication framework for metaverse application (GSCM) to explore and define semantic information through the goal levels. Specifically, we first analyze the traditional image communication framework in metaverse construction and then detail our proposed semantic information along with the end-to-end wireless communication. We then describe the designed modules of the GSCM framework, including goal-oriented semantic information extraction, base knowledge definition, and neural radiance field (NeRF) based metaverse construction. Finally, numerous experiments have been conducted to demonstrate that, compared to image communication, our proposed GSCM framework decreases transmission latency by up to 92.6% and enhances the virtual object operation accuracy and metaverse construction clearance by up to 45.6% and 44.7%, respectively.

As a key technology in beyond 5G and future 6G communications, RAN slicing can realize differentiated service level agreement (SLA) guarantees. In this paper, we investigate the multi-slice multi-user RAN slicing. A dynamic bandwidth allocation scheme for RAN slicing is proposed based on hierarchical intelligence, where bandwidth pre-allocation and hyper-parameter tuning for MAC layer schedulers are jointly optimized to maximize the system utility, i.e., the weighted sum of spectrum efficiency (SE) and SLA satisfaction ratio (SSR) of different slices. The problem is formulated as a twin-time scale Markov decision process (MDP), where the bandwidth pre-allocation and the scheduler parameter tuning are performed on a long-term scale (e.g., seconds) and a short-term scale (e.g., 100 ms), respectively, for which we propose a hierarchical twin-time scale Dueling deep Q learning network (TTS-DDQN) algorithm. A new reward-clipping mechanism is proposed to get a better trade-off between stabilized training and higher system utility. In order to improve the robustness to time-varying traffic patterns and non-stationary dynamic environments, we further propose a traffic-aware module for more efficient sampling of the experience pool, and a variational adversarial inverse reinforcement learning (VAIRL) module for reward automation design. Extensive simulations show that the traffic-aware TTS-DDQN in stationary scenarios and the VAIRL module embedded TTS-DDQN in non-stationary scenarios outperform existing typical DQN-based algorithms, hard slicing and non-slicing, etc.

This paper studies the computational offloading of CNN inference in dynamic multi-access edge computing (MEC) networks. To address the uncertainties in communication time and edge servers’ available capacity, we propose a novel semantic compression method, autoencoder-based CNN architecture (AECNN), for effective semantic extraction and compression in partial offloading. In the semantic encoder, we introduce a feature compression module based on the channel attention mechanism in CNNs, to compress intermediate data by selecting the most informative features. Additionally, to further reduce communication overhead, we leverage entropy encoding to remove the statistical redundancy in the compressed data. In the semantic decoder, we design a lightweight decoder to reconstruct the intermediate data through learning from the received compressed data to improve accuracy. To effectively trade-off communication, computation, and inference accuracy, we design a reward function and formulate the offloading problem of CNN inference as a maximization problem with the goal of maximizing the average inference accuracy and throughput over the long term. To address this maximization problem, we propose a graph reinforcement learning-based AECNN (GRL-AECNN) method, which outperforms existing works DROO-AECNN, GRL-BottleNet++ and GRL-DeepJSCC under different dynamic scenarios. This highlights the advantages of GRL-AECNN in offloading decision-making for CNN inference tasks in dynamic MEC.

This paper studies the computational offloading of CNN inference in device-edge co-inference systems. Inspired by the emerging paradigm semantic communication, we propose a novel autoencoder-based CNN architecture (AECNN), for effective feature extraction at end-device. We design a feature compression module based on the channel attention method in CNN, to compress the intermediate data by selecting the most important features. To further reduce communication overhead, we can use entropy encoding to remove the statistical redundancy in the compressed data. At the receiver, we design a lightweight decoder to reconstruct the intermediate data through learning from the received compressed data to improve accuracy. To fasten the convergence, we use a step-by-step approach to train the neural networks obtained based on ResNet-50 architecture. Experimental results show that AECNN can compress the intermediate data by more than 256 × with only about 4% accuracy loss, which outperforms the state-of-the-art work, BottleNet++. Compared to offloading inference task directly to edge server, AECNN can complete inference task earlier, in particular, under poor wireless channel condition, which highlights the effectiveness of AECNN in guaranteeing higher accuracy within time constraint.

This paper studies the computational offloading of CNN inference in dynamic multi-access edge computing (MEC) networks. To address the uncertainties in communication time and Edge servers' available capacity, we use early-exit mechanism to terminate the computation earlier to meet the deadline of inference tasks. We design a reward function to trade off the communication, computation and inference accuracy, and formu-late the offloading problem of CNN inference as a maximization problem with the goal of maximizing the average inference accuracy and throughput in long term. To solve the maxi-mization problem, we propose a graph reinforcement learning-based early-exit mechanism (GRLE), which outperforms the state-of-the-art work, deep reinforcement learning-based online offloading (DROO) and its enhanced method, DROO with early-exit mechanism (DROOE), under different dynamic scenarios. The experimental results show that G RLE achieves the average accuracy up to 3.41 x over graph reinforcement learning (GRL) and 1.45x over DROOE, which shows the advantages of GRLE for offloading decision-making in dynamic MEC.

This paper studies inference acceleration using distributed convolutional neural networks (CNNs) in collaborative edge computing. To ensure inference accuracy in inference task partitioning, we consider the receptive-field when performing segment-based partitioning. To maximize the parallelization between the communication and computing processes, thereby minimizing the total inference time of an inference task, we design a novel task collaboration scheme in which the overlapping zone of the sub-tasks on secondary edge servers (ESs) is executed on the host ES, named as HALP. We further extend HALP to the scenario of multiple tasks. Experimental results show that HALP can accelerate CNN inference in VGG-16 by 1.7-2.0x for a single task and 1.7-1.8x for 4 tasks per batch on GTX 1080TI and JETSON AGX Xavier, which outperforms the state-of-the-art work MoDNN. Moreover, we evaluate the service reliability under time-variant channel, which shows that HALP is an effective solution to ensure high service reliability with strict service deadline.

This paper studies inference acceleration using distributed convolutional neural networks (CNNs) in collaborative edge computing network. To avoid inference accuracy loss in inference task partitioning, we propose receptive field-based segmentation (RFS). To reduce the computation time and communication overhead, we propose a novel collaborative edge computing using fused-layer parallelization to partition a CNN model into multiple blocks of convolutional layers. In this scheme, the collaborative edge servers (ESs) only need to exchange small fraction of the sub-outputs after computing each fused block. In addition, to find the optimal solution of partitioning a CNN model into multiple blocks, we use dynamic programming, named as dynamic programming for fused-layer parallelization (DPFP). The experimental results show that DPFP can accelerate inference of VGG-16 up to 73% compared with the pre-trained model, which outperforms the existing work MoDNN in all tested scenarios. Moreover, we evaluate the service reliability of DPFP under time-variant channel, which shows that DPFP is an effective solution to ensure high service reliability with strict service deadline.

This paper studies inference acceleration using distributed CNNs in collaborative edge computing network. To ensure no inference accuracy loss in task partitioning, we propose receptive field-based segmentation. To reduce the computation time and communication overhead, we propose a novel collaborative edge computing using fused-layer parallelization to partition a CNN model into multiple blocks. To find the optimal partition of a CNN model, we use dynamic programming, named as DPFP. To address computation heterogeneity of edge servers (ESs), we design a low-complexity search algorithm which can select the optimal subset of collaborative ESs for inference. The experimental results show that DPFP can accelerate inference up to 71% for ResNet-50 and 73% for VGG-16 compared to running the pre-trained models, which outperforms the existing works MoDNN and DeepSlicing. Moreover, we propose an analytical method to estimate the speedup ratio of different GPU platforms by using FLOPs and effective computing capacity. Furthermore, we evaluate the service failure probability under time-variant channel and variation of image sizes, which shows that DPFP is effective to ensure high service reliability with strict service deadline.

The main challenge of single image super resolution (SISR) is the recovery of high frequency details such as tiny textures. However, most of the state-of-the-art methods lack specific modules to identify high frequency areas, causing the output image to be blurred. We propose an attention-based approach to give a discrimination between texture areas and smooth areas. After the positions of high frequency details are located, high frequency compensation is carried out. This approach can incorporate with previously proposed SISR networks. By providing high frequency enhancement, better performance and visual effect are achieved. We also propose our own SISR network composed of DenseRes blocks. The block provides an effective way to combine the low level features and high level features. Extensive benchmark evaluation shows that our proposed method achieves significant improvement over the state-of-the-art works in SISR.

This paper considers the problem of transmission scheduling of delay-sensitive data over a point-to-point correlated Rayleigh fading channel with channel estimation errors. According to the imperfect channel state information (CSI) and the buffer state, the transmit power and the modulation and coding scheme (MCS) are determined to jointly maximize the energy efficiency, and minimize transmission delay and overflow probability. To account of the effects of the channel estimation errors, the CSI imperfection is modeled as uncertain sets using the ellipsoidal approximation. Then the joint optimization problem is formulated using the weighted sum method. Using the idea of online learning, two algorithms are proposed to schedule the delay-sensitive data for the situations with and without the uncertainty bound of channel estimation, respectively. The numerical results indicate that the proposed online learning based scheduling algorithms can tackle the imperfect CSI issue and improve the system performance in terms of the energy efficiency, transmission delay and overflow probability. Moreover, the convergence times are very short, which highlights the feasibility of the proposed online learning based scheduling for practical systems.