The growing integration of renewable energy in the power system contributes to achieving the goal of carbon neutrality but brings a critical challenge to frequency stability. With the development of vehicle-to-grid, electric vehicles (EV) as flexible energy storage systems can provide stable frequency regulation services for the power system. However, synchronizing state messages from distributed EVs to virtual power plants (VPP) generates numerous latency-sensitive and computation-intensive tasks that make providing efficient frequency regulation services challenging. Therefore, this paper proposes a joint computation offloading and frequency regulation bidding scheme for EVs. First, the cloud-edge collaboration framework is applied in VPP to achieve efficient management of large-scale EVs. Then, we formulate frequency regulation as a joint optimization problem that balances service revenue against computational resource costs. Using Lyapunov optimization, long-term delay constraint in the optimization problem is transformed into a queue stability requirement to simplify the problem. The deep Q-learning algorithm is then employed to solve the optimal computation offloading strategies, and the Lagrange multiplier method determines real-time frequency regulation bidding for dynamically updated EV clusters. Extensive simulations demonstrate that the proposed scheme can substantially reduce computing cost and improve frequency regulation profit up to 30%.

Network slicing provides personalized services by building isolated logical networks. Besides, through deploying Virtual Network Function (VNF) chains on Mobile Edge Computing (MEC) servers, network slicing can guarantees the performance of low-latency services. To reduce maintenance costs, it is necessary to optimize the deployment of both the slices’ access functions on the base stations and the VNF chains on the MEC servers. Meanwhile, optimizing MEC server activation can save energy while meeting service demands. There have been many studies focusing on MEC server activation, access function deployment, and VNF chain deployment, but no work has considered them simultaneously. To fill this research gap, we construct a joint server activation as well as the access functions and VNF chains deployment model, aiming to guarantee performance while minimizing the cost. This problem is a mixed-integer nonlinear programming problem which is hard to be solved. Thus, we decompose the problem and design a Deep Q Network (DQN)based three-stage algorithm. Numerical results demonstrate the superiority of the proposed scheme over the baseline methods.

Antenna tuning plays an essential role in ensuring high quality wireless communications. Targeting for higher Quality of Service (QoS), many existing network antenna tuning schemes are based on expert knowledge, rule-based policies or conventional optimization theory. However, maximizing the traffic-related QoS does not guarantee that all customers experience good services. In addition, existing schemes are often limited to some handcrafted rules or heuristics and lack of adaptability especially in a time-varying environment. Quality of Experience (QoE), a user-centric metric, can better measure users' satisfaction for services in wireless networks. This paper proposes the cooperative tuning of antennas based on QoE, a paradigm shift from network-centric QoS to user-centric QoE domain. In a normal cellular network, besides the need of improving the overall QoE, it requires handling faults from different cells. As Multi-agent Reinforcement Learning (MARL) has the capability of self-learning the dynamics of environment, we propose an antenna configuration algorithm based on multi-goal MARL. In our framework, agents from different cells not only need to cooperate with each other to achieve the global goal of increasing the overall QoE of the wireless network but also complete some personal goals by combating the faults encountered in their own cells. To accelerate the training efficiency, we introduce a novel two-stage curriculum learning. To reduce the collection time of each QoE sample, we develop an accurate and timely QoE/QoS mapping model with the cascading of a Random Forest Classifier (RFC) and a Deep Neural Network (DNN) (abbreviated as RFC-DNN), which can help us obtain QoE by collecting QoS measurements and perform QoE-based antenna configurations with smaller time granularity. Our proposed RFC-DNN model can reduce the time by 70% when predicting the QoE of a single sample. A huge amount of time will be saved in MARL when tens of thousands of transitions/samples need to be collected. The performance results show that our proposed antenna tuning schemes can not only address specific faults in each cell, but also significantly improve the global average QoE with a faster and more stable convergence speed.

The slice-based Open Radio Access Network (O-RAN) enables rapid deployment of logical networks and provides personalized services. In designing the inter-slice resource sharing scheme, multiple factors need to be considered. Firstly, resource isolation level (RIL) and interference isolation level (IIL) are two key factors in ensuring slice isolation. Secondly, the new business models brought by network slicing imply the need to consider tenants’ resource costs. However, to date, no prior studies have considered RIL, IIL and costs simultaneously, often involving only one or two of them. Accordingly, we propose an inter-slice resource block (RB) leasing and association adjustment scheme (ISRLA) that allows slices to borrow RBs from the shared resource pool and other slices. ISRLA is divided into three modules, namely, prediction module, leasing module, and RB association adjustment module. Firstly, the prediction module uses a Long Short-Term Memory model to predict RB demand and guide subsequent resource optimization. Then, the leasing module balances RIL and costs to determine the number of RBs to be leased in or leased out. This is defined as a nonlinear integer programming problem, which is solved through an iterative strategy based on camp swapping (ISBCS). ISBCS is further proven to be convergent. Finally, the RB association adjustment module determines specific RB adjustment strategies based on the results obtained from the leasing module, aiming to ensure IIL. Here, we design a potential game based on the Lagrangian relaxation (LR) method to obtain an approximate optimal solution while reducing computational complexity. Compared with camp enumeration (CE), solver (MOSEK), and genetic algorithm (GA), the proposed ISRLA reduces simulation time by up to 84%, 95%, and 99%, respectively. When the number of slices is 4, the performance of ISRLA is the same as that of CE. Moreover, the IIL gap between ISRLA and MOSEK is less than 4.5%. The simulation results show that the proposed ISRLA can obtain the approximate optimal solution with low computational complexity.

Cellular traffic prediction is crucial for intelligent network operations, such as load-aware resource management and proactive network optimization. In this paper, to explicitly characterize the temporal dependence and spatial relationship of nonstationary real-world cellular traffic, we propose a novel prediction method. First, we decompose traffic data into three components which represent various cellular traffic patterns. Second, to capture the spatial relationship among base stations (BSs), we model each component as a directed causal graph by variable-lag transfer entropy (VLTE) based causal structure learning. Third, we design a deep learning model combining graph attention network (GAT) and gated recurrent unit (GRU) to predict each component. GRU is used to capture temporal dependence. GAT is trained to quantitatively analyze spatial relationship and aggregate spatial features. Finally, we integrate the prediction results of three components to obtain the cellular traffic prediction result. We conduct extensive experiments on real-world traffic data, and the results show that our proposed method outperforms other common methods.

Relay cooperation with energy harvesting provides a promising solution to alleviate coverage limitations and energy constraints in hybrid visible light communication (VLC)/radio frequency (RF) Internet of Things (IoT) system with simultaneous lightwave information and power transfer (SLIPT). In the hybrid system, the VLC service provider (VLCSP) seeks the cooperation of relay nodes (RNs) for information delivery to a certain end node (EN). However, considering the autonomous behaviors of the RNs, there are two challenging issues to address in facilitating relay cooperation: 1) selfishness and 2) information asymmetry. Hence, this article proposes a novel incentive scheme for relay cooperation based on an agency selling format. Unlike previous incentive scheme designs, first, the VLCSP charges the cooperating RN for energy harvesting per unit energy price. Second, once the information is successfully transmitted to the EN, the VLCSP pays a portion of future revenue as an agency payment to the RN, in a format seen as agency selling. By giving the VLCSP pricing power, it needs to design a mutually agreeable contract that includes a menu of unit energy prices and agency payments. Here, aiming to maximize the VLCSP’s expected utility, we apply a joint adverse selection and moral hazard model to formulate and optimize the contract design problem in the presence of information asymmetry. Then, the optimal contract solution is derived by using Lagrangian dual analysis. Besides, we present two extreme scenarios where only adverse selection or moral hazard model is applied. Numerical results illustrate that the proposed incentive scheme outperforms the considered benchmarks in terms of the expected utility of VLCSP and RNs, as well as the expected social welfare. We also demonstrate the incentive efficiency of our scheme

With the development of telecommunication technology and the popularity of intelligent devices, user experience in cellular network has become the primary factor of concern. At the same time, network operators are plagued by two problems: Prediction of users real-time experience and find network parameters that have a decisive impact on user experience. We proposed a novel scheme for user experience prediction to deal with these two problems. Causal structure learning for cellular networks was used to analyze numerous performance indicators (KPIs) collected from base stations and key quality indicators (KQIs). Through causal structure learning, a directed causal graph based on the association between KPIs and KQI can be obtained. This causal structure can be embedded in graph attention network. Among them, attention mechanism was selected to further strengthen the correlation between parameters. This correlation between each KPI and between KPIs and KQI was used to predict future value of cell level user experience. Results showed that proposed method performance well in cellular network data analysis and user experience prediction.

Millimeter-wave (mmWave) communication has been regarded as one of the most promising means to improve system throughput in the fifth-generation (5G) era. However, there exists signal blockage problem in mmWave network. We consider the combination of mmWave communication with ultradense network (UDN) and virtual cell structure. UDN allows multiple small base stations (SBSs) to be deployed within a limited area. user-centric virtual cell structure enables users to associate multiple mmWave SBSs simultaneously by using coordinated multi-point (CoMP) transmission technology. Therefore, we formulate user association problem for ultra-dense mmWave networks to maximize system throughput while guaranteeing quality of service (QoS) requirements of users. To solve this nonlinear binary integer programming optimization problem, a novel user association scheme based on the many - to-many two-sided matching game with externalities is proposed. To characterize the properties of proposed scheme, we prove that it converges to the two-sided exchange stability within a limited number of iterations. Simulation results show that proposed scheme significantly outperforms other schemes in improving user rate and system throughput while guaranteeing QoS requirements of users.