As a new communications architecture, the Space-Air-Ground integrated network (SAGIN) integrates satellites, airborne platforms, and terrestrial networks to enhance global connectivity and support robust and flexible communication capabilities. Efficient task offloading and resource allocation are crucial for SAGIN to meet the quality of service (QoS) requirements at low cost. In this paper, we formulate task offloading and resource allocation as a time-sequential decision-making problem, aiming to maximize task completion within available communication and computational resources. We propose an online approach referred to as graph fusion deep reinforcement learning (GF-DRL). GF-DRL incorporates a graph feature extraction network that utilizes a graph convolutional network (GCN) to extract features from both the task graph and user equipment (UE) graph, along with two attention mechanisms (hard and soft) to merge the two graphs. We also propose an action encoding and mapping network to generate both discrete (offloading) and continuous (allocation) decisions in an end-to-end manner. Simulation results validate the effectiveness of our proposed GF-DRL compared to state-of-the-art task offloading resource allocation approaches.

Space-Air-Ground integrated network (SAGIN) is a crucial component of the 6G, enabling global and seamless communication coverage. This multi-layered communication system integrates space, air, and terrestrial segments, each with computational capability, and also serves as a ubiquitous computing platform. An efficient task offloading and resource allocation scheme is key in SAGIN to maximize resource utilization efficiency, meeting the stringent quality of service (QoS) requirements for different service types. In this paper, we introduce a dynamic SAGIN model featuring diverse antenna configurations, two timescale types, different channel models for each segment, and dual service types. We formulate a problem of sequential decision-making task offloading and resource allocation. Our proposed solution is an innovative online approach referred to as graphic deep reinforcement learning (GDRL). This approach utilizes a graph neural network (GNN)-based feature extraction network to identify the inherent dependencies within the graphical structure of the states. We design an action mapping network with an encoding scheme for end-to-end generation of task offloading and resource allocation decisions. Additionally, we incorporate meta-learning into GDRL to swiftly adapt to rapid changes in key parameters of the SAGIN environment, significantly reducing online deployment complexity. Simulation results validate that our proposed GDRL significantly outperforms state-of-the-art DRL approaches by achieving the highest reward and lowest overall latency.

Space-Air-Ground integrated network (SAGIN) is a crucial component of the 6G, enabling global and seamless communication coverage. An efficient task offloading and resource allocation scheme is key in SAGIN to maximize resource utilization efficiency, meeting the stringent quality of service (QoS) requirements for different service types. In this paper, we introduce a dynamic SAGIN model featuring diverse antenna configurations, two timescale types, different channel models for each segment, and dual service types. We formulate a problem of sequential decision-making task offloading and resource allocation. Our proposed solution is an innovative online approach referred to as graphic deep reinforcement learning (GDRL). This approach utilizes a graph neural network (GNN)-based feature extraction network to identify the inherent dependencies within the graphical structure of the states. Simulation results validate that our proposed GDRL significantly outperforms state-of-the-art deep reinforcement learning (DRL) approaches by achieving the highest reward and lowest overall latency.

Network slicing is a key enabler of 5G and beyond networks to satisfy the diverse quality of service (QoS) requirements of different services simultaneously. In network slicing, radio access network (RAN) slicing is essential to establish a functional network slice by connecting mobile devices and mapping virtualized resource units to different slices. This requires a highly efficient resource allocation scheme to maximize resource utilization efficiency and meet the diverse QoS requirements. In this paper, we propose a dynamic RAN slicing model that incorporates multiple distributions to accommodate different user request types and diverse priorities among traffic types in the same slice, where the total available resources are dynamically changing over time. We formulate resource allocation as a time-sequential dynamic optimization problem that takes into account system stability, resource limitation, different timescales, long-term system performance, and user priority. We propose a deep reinforcement learning-based (DRL-based) approach referred to as prediction-aided weighted DRL (PW-DRL) to online infer the power allocation and user acceptance decisions that can maximize a predefined reward function. Additionally, a prediction network is formulated to capture the correlation between current and future states. Simulation results validate that our proposed PW-DRL significantly outperforms state-of-the-art approaches by achieving the highest long-term reward and fastest convergence.

Network slicing is key to enabling 6G and beyond networks to simultaneously meet the diverse quality of service (QoS) requirements of various services. In network slicing, radio access network (RAN) slicing is essential to establish a functional network slice by connecting mobile devices and mapping virtualized resource units to different slices. This demands an efficient resource allocation scheme that maximizes resource utilization while meeting diverse QoS requirements. In this paper, we propose a new dynamic resource allocation framework that encompasses three types of services. We formulate a dynamic resource allocation problem that features a mixed action space and has both long-term power and instantaneously available resource unit constraints. We propose a deep reinforcement learning (DRL)-based approach referred to as prediction-aided weighted DRL (PW-DRL), which infers the power allocation and user acceptance decisions to maximize a predefined reward function. Additionally, we propose a prediction network that significantly improves the DRL learning process under limited resources by supplying future state information. Simulation results validate that our proposed PW-DRL significantly outperforms state-of-art DRL approaches by achieving the highest long-term reward and fastest convergence.

Mobile traffic prediction opens a promising avenue to demand-aware large-scale resource allocation with a significant improvement in the spectral efficiency. Various long-term prediction methods have been proposed in the literature. However, when considering the stringent requirement of the real-time and efficient radio resource allocation for future wireless communications, developing short-term prediction methods with high prediction accuracy is more desirable. In this paper, we exploit spatiotemporal correlations among the mobile traffic data and propose a novel machine learning-based short-term prediction method, referred to as spatiotemporal Gaussian Process Kalman filter (ST-GPKL) method, which includes two phases: the model selection and inference. The function of the model selection is to fine-tune the hyperparameters of the designed kernel function, while that of the inference incorporates the Kalman filter to predict the future mobile data traffic. Compared with the conventional methods, the proposed one can significantly improve the prediction accuracy, resulting in much higher efficiency in large-scale resource allocation.