Offloading computationally intensive tasks to edge nodes reduces latency and improves user experience. Unmanned aerial vehicle (UAV)-assisted multi-access edge computing (MEC) can effectively address the limitations of the fixed deployment of traditional edge nodes, but the dynamic nature of UAVs brings challenges to the optimization of computation offloading. Methods based on deep reinforcement learning (DRL) can efficiently learn edge network dynamics to optimize computation offloading in UAV-assisted systems. In the system model, different edge nodes such as UAVs, roadside units, and user equipments (UEs) are treated as agents in the reinforcement learning network, so that the optimization of the edge computation offloading strategy can be transformed into a multi-agent optimization problem. In addition, meta-reinforcement learning is introduced into the multi-agent deep deterministic policy gradient (MADDPG) algorithm to cope with the dynamics and uncertainty of the UAV-assisted edge computing network environment. Simulation results show that the computation offloading strategy based on meta-reinforcement learning can not only quickly adapt to the dynamic changes of the network environment but also outperform other benchmark algorithms in terms of network energy efficiency.

The composition of soundtracks for videos is a highly integrated process that synthesizes artistic creativity with technical expertise. It involves the precise synchronization of musical elements with the visual component to effectively enhance the artistic quality of the video. Herein, we aimed to generate harmonious and temporally aligned music based on video content, assisting artists in soundtrack composition. Existing approaches predominantly generate music based on human motion data from fixed-length videos, limiting their applicability. To overcome this limitation, we propose an autoregressive unit diffusion model,$\mathbf {W^{2}}$, for variable-length video-to-music generation.$\mathbf {W^{2}}$segments videos and music into unified temporal units and employs a diffusion model to generate music units conditioned on structured video information and contextual music information. It then autoregressively generates coherent and consistent music sequences. This model employs structured video condition information, obtained through global encoding, to effectively generate music units that consistently align with the video content. At the same time, contextual music condition information is used to ensure the generation of a continuous sequence of music units matching the duration of the video. Unlike prior models relying on human skeletal features,$\mathbf {W^{2}}$can be applied to not only dance videos but also free-style videos. Experimental results show that$\mathbf {W^{2}}$generates music that is temporally synchronized and semantically aligned with visual content.

Multi-access edge computing has become one of the key technologies involved in the development of the Internet of Vehicles (IoV) because of its low latency and high bandwidth. In some hot spots or emergency situations, unmanned aerial vehicle (UAV)-assisted vehicular edge computing can flexibly cope with the problem of insufficient resources in a fixed edge server network. Deep reinforcement learning (DRL) can effectively address the edge computing offloading optimisation problem in the above scenarios and improve the utilisation of network resources. In this paper, we transform the optimisation problem of a UAV-assisted edge computing offloading strategy in an IoV context into a multi-agent optimisation problem by establishing a system model. In addition, a graph attention network (GAT) is introduced to determine the interaction relationships between multiple agents, such as multiple UAVs and edge servers. The deep deterministic policy gradient (DDPG) algorithm combined with a GAT can effectively capture the collaboration pattern between multiple agents to obtain the computational offloading strategy that minimises the system delay. The simulation results show that the convergence speed and optimisation ability of the proposed algorithm based on the GAT-DDPG fusion network are superior to those of other baseline algorithms.

With the rapid development of new era satellite communication technology, content delivery network (CDN) using satellite is a paradigm to build intelligent distribution networks. However, how to improve the performance of the edge and cloud resources is still a challenge when satellites perform CDN distribution. Considering that the problem is a non-convex mixed-integer optimization problem, we propose a greedy strategy-deep deterministic policy gradient (GS-DDPG) algorithm by combining deep reinforcement learning with traditional optimization algorithms. A trial-reward feedback mechanism is established to accumulate experience and learn the caching policy of the network. The algorithm aims to minimize delay and transmission energy consumption, which optimizes the edge-cloud collaborative caching decision and satellite resource scheduling problem in the satellite CDN distribution task. Simulation results show that the proposed GS-DDPG algorithm can effectively improve the system efficiency in different network environments.

With the rapidly growing demand for mobile edge computing (MEC), unmanned aerial vehicles (UAVs) are increasingly used in task offloading and area coverage. However, high energy consumption and computational complexity become the key challenges limiting their development. In this paper, we propose the tiny multi-agent deep deterministic policy gradient (Tiny-MADDPG) algorithm based on deep reinforcement learning (DRL), which aims to optimize the energy consumption of multi-UAV-assisted MEC systems. By constructing a multi-agent deep deterministic policy gradient (MADDPG) framework and combining it with a lightweight network design, the algorithm achieves joint optimization of flight paths, task offloading and resource allocation. For the multi-UAV collaboration scenario, a comprehensive reward function containing energy consumption, delay and coverage is designed to ensure the overall improvement of system performance. Simulation experiments show that TinyMADDPG achieves significant results in minimizing energy consumption compared with other benchmark algorithms. The results validate the dual advantages of multi-agent collaboration and lightweight design, and provide a valuable reference for energy consumption optimization of multi-UAV systems.

Offloading computing tasks to edge servers can provide better user experience. However, the deployment of a large number of distributed edge nodes brings challenges to the optimal management of network energy consumption. In this paper, deep reinforcement learning (DRL) is used to optimize unmanned aerial vehicle (UAV)-assisted edge computing task offloading to improve network energy efficiency. First, different edge nodes are treated as agents in the DRL network, so that the optimization of edge computing task offloading strategy is transformed into a multi-agent optimization problem. In addition, meta-learning is introduced into the multi-agent deep deterministic policy gradient algorithm to cope with the dynamics and uncertainty of the edge computing network environment. The simulation results show that the task offloading strategy based on meta-learning can not only quickly adapts to the dynamic changes of the environment and tasks, but also outperforms the benchmark algorithms in network energy efficiency.

With the help of unmanned aerial vehicles (UAVs), mobile ultrahigh definition (UHD) live broadcast systems can not only provide UHD images from multiple angles, but also monitor situations in live broadcast scenes in real time, thus providing a better user experience. However, owing to the restricted flight and edge computing capabilities of UAVs, how to optimise the assignment strategy when multiple UAVs perform tasks collaboratively becomes pivotal for system performance. In this paper, we optimise the multi-UAV collaborative task assignment strategy by improving the wolf pack algorithm (WPA). The strategy incorporates the ideas of crossover, replication, and mutation in genetic algorithms in the process of position updating, and improves the update mode of new population individuals. In addition, the strategy introduces the idea of auction algorithm to correct the infeasible solution that violates the constraints to obtain the optimal strategy. The simulation results show that the proposed algorithm can effectively solve the collaborative task assignment problem with better stability and convergence.

Multimodal sentiment analysis (MSA) is important for quickly and accurately understanding people's attitudes and opinions about an event. However, existing sentiment analysis methods suffer from the dominant contribution of text modality in the dataset; this is called text dominance. In this context, we emphasize that weakening the dominant role of text modality is important for MSA tasks. To solve the above two problems, from the perspective of datasets, we first propose the Chinese multimodal opinion-level sentiment intensity (CMOSI) dataset. Three different versions of the dataset were constructed: manually proofreading subtitles, generating subtitles using machine speech transcription, and generating subtitles using human cross-language translation. The latter two versions radically weaken the dominant role of the textual model. We randomly collected 144 real videos from the Bilibili video site and manually edited 2557 clips containing emotions from them. From the perspective of network modeling, we propose a multimodal semantic enhancement network (MSEN) based on a multiheaded attention mechanism by taking advantage of the multiple versions of the CMOSI dataset. Experiments with our proposed CMOSI show that the network performs best with the text-unweakened version of the dataset. The loss of performance is minimal on both versions of the text-weakened dataset, indicating that our network can fully exploit the latent semantics in nontext patterns. In addition, we conducted model generalization experiments with MSEN on MOSI, MOSEI, and CH-SIMS datasets, and the results show that our approach is also very competitive and has good cross-language robustness.

Unmanned-aerial-vehicle (UAV)-assisted multiaccess edge computing (MEC) networks can effectively broaden the application scope of the Internet of Things (IoT) in complex scenarios, such as maritime operations, military communications, and emergency commands. However, uncertain factors, such as weather changes and temporary airspace control, pose great challenges to UAV flight safety. Obstacles resulting from these uncertain factors may intersect with UAVs with preplanned flight paths, leading to accidents. Therefore, generating the optimal flight trajectory to avoid these obstacles is key in the successful operation of this fuzzy system. In this article, we present a heuristic trajectory generation scheme for complex offshore environments that can generate optimal trajectories according to complex terrain conditions and avoid uncertain obstacles. First, we build a complex terrain model based on a 3-D offshore environment to simulate the conditions in UAV-assisted MEC networks. Second, we propose a network performance optimization objective function that is based on UAV characteristics. Third, we improve the existing ant colony optimization (ACO) algorithm by introducing chaotic mapping, polarizing the pheromone recording rule, and implementing a simulated annealing screening mechanism to efficiently generate trajectories. Finally, we design an efficient obstacle avoidance algorithm for different combinations of obstacle regions. The simulation results show that our proposed trajectory generation scheme can efficiently avoid obstacles and significantly improve the total trajectory loss rate compared with that of baseline schemes.

Due to the development of social media, the threshold for information dissemination has become lower than ever before. As a special kind of information, rumors are usually harmful and are usually accompanied by a high degree of ambiguity that makes them difficult to immediately identify, but “rumors stop at wise men.” When someone identifies a rumor as false and begins spreading the truth instead, a confrontational relationship obtains between the rumor and the truth that leads to the stifling of the former. Given this, we developed a 2SIH2R model in this study that contains mechanisms of discernment and confrontation in a heterogeneous network to examine the dissemination of the rumor and the truth. By using mean‐field equations of the 2SIH2R model, the threshold of the spreading of each can be determined separately in three cases. The results of a numerical simulation show that under the same conditions, the greater is the mechanism of discernment or confrontation, the smaller is the instantaneous maximum influence and the final range of influence of the rumor. It can be also concluded that the earlier release of the truth about the event by the government can significantly control the rumor. Secondly, it is more effective to publish the truth in advance than after the rumor has appeared. Thirdly, it is more important for the government to increase education and improve the ability of citizens to reveal the rumor than to increase the spread of the truth after the rumor occurs. These results can be used to help reduce the harmful effects of rumors.

With human society stepping into the data era, deep learning has been widely used in various industries. However, in the training process of deep learning, underfitting and overfitting are often encountered, leading to poor network generalization performance. Based on a Convolutional Neural network (CNN), this paper optimizes the model by mitigating underfitting and overfitting. Incorporating multiple approaches, the accuracy of the model is finally improved by 4 percentage points by adjusting the learning rate and adding regularization, etc.

Mobile users can be profiled through network data analysis. The classification of subway passengers will help improve operating efficiency and safety. Here, we investigate how to classify mobile service users using subway networks. We not only discuss mobile subway user data analytics in general but also present the entire process of wireless data collection and big data analytics. We first introduce the characteristics of external data recording (XDR) collected from deep packet inspection (DPI) device in current wireless networks. We then propose an algorithm to distinguish subway residents from commuters. Finally, we examine the use and performance of several machine learning algorithms to classify spatio-temporal and service usage patterns of mobile users. Through the accurate classification of users, operators can provide personalized service.

The rapid growth in the energy consumption of mobile networks has become a major concern for mobile operators. Today's mobile networks' usage is dominated by over-the-top (OTT) applications, and operators are keen to determine the network energy consumed by these OTT applications. With a recent shift in user behavior toward a preference for instant messaging (IM) applications over conventional mobile services, operators are interested in exploring what impact OTT IM applications such as WeChat will have on the energy consumption of a network when compared with a corresponding conventional mobile service. Here, we present for the first time energy assessment models for mobile services based on real network and service measurements to address this need. Using WeChat as an OTT IM application example, our results show that WeChat consumes more network energy than conventional mobile services for both light users and heavy text users due to the network signaling energy overhead. In comparison, for heavy voice users, WeChat consumes less network energy since voice messages are first recorded and then sent in packet bursts. Our findings provide a quantitative analysis of the energy consumption of mobile services, which should be valuable for mobile operators and OTT application developers to improve the energy-efficiency of mobile applications and services.