In vehicular cooperative perception (CP), numerous resource allocation strategies have been proposed to enhance urban autonomous driving. However, existing studies often overlook the competition between self-perception and cooperative perception, where degrading a ground vehicle's (GV's) self-perception may introduce safety risks and reduce passenger comfort. Moreover, UAV–GV cooperation—which can improve sensing precision, reduce task execution delay, and enhance CP service availability—has received limited attention. It is worth noting that unmanned aerial vehicles (UAVs) are unavailable for cooperative perception during the recharging process. To address these issues, this paper investigates on-demand scheduling strategy in UAV–GV cooperative perception. At the millisecond timescale, resource competition is considered in real-time sensing, communication, and computation (SC2) resource allocation. At the minute timescale, the idle flying period between consecutive tasks is utilized for UAV recharging through attachment to GVs along the route. Specifically, we first develop a model that captures the mutual influence between UAVs and GVs on perception performance under resource constraints. Then, a mixed-timescale solution is proposed: at the small timescale, a multi-agent deep reinforcement learning (MA-DRL) algorithm with gradient-free projection and auxiliary supervision is designed to schedule SC2resources; at the large timescale, a Hungarian-based algorithm is employed to control UAV recharging. Simulation results show that the proposed approach outperforms benchmark schemes by reducing task execution delay and energy consumption, and enhancing CP service availability, while satisfying sensing precision, GV safety, and passenger comfort requirements.

In vehicular edge computing networks, the realtime task processing for different vehicles by determining task offloading paths and allocating offloading resources is crucial. However, existing studies often overlook the impact of air-ground cooperation, the task burden on offloading relay nodes, and the limited resources on network performance. These oversights can lead to traffic congestion and performance degradation. To address these issues, this paper proposes a novel integrated sensing and communication (ISAC)-enabled unmanned aerial vehicle (UAV)-assisted task offloading strategy, where tasks are relayed and processed by ground vehicles and multiple UAVs based on the task load rate. Specifically, we first analyze the differences between offloading paths for ground vehicles and UAVs, the coupling relationship between communication, sensing, and computing resources, and the conflicts between UAV self-sensing and ground vehicle offloading assistance. We then formulate the offloading path determination and resource allocation problem as an optimization problem, to minimize overall offloading delay and UAV energy consumption. Subsequently, we propose a novel offloading algorithm based on deep reinforcement learning to schedule offloading paths and network resources intelligently. Finally, simulation results demonstrate the effectiveness of our proposal in reducing offloading delay and UAV energy consumption.

Vehicular cooperative perception (VCP) facilitates the exchange of sensing data among vehicles through vehicle-to-everything (V2X) communication, significantly increasing the sensing range and precision of individual autonomous vehicles (AVs). However, efficiently managing the sharing and processing of large volumes of sensing data presents challenges due to restricted communication and computation resources. This study introduces an integrated sensing, communication, and computation (ISCC)-based task offloading and resource allocation (ITORA) framework, which optimizes cooperative perception by determining what data to share, which vehicles to involve, and how to process the data effectively. We develop an information value function to evaluate the data quality for each vehicle. Subsequently, we design strategies for sensing task allocation, task offloading, and resource allocation to enable value-driven data selection at a subregion level, facilitating collaborative computing among edge servers and vehicles. Additionally, we formulate an optimization problem aimed at maximizing information value while minimizing delay and energy consumption, subject to constraints on a full region of interest (RoI) coverage, delay, wireless bandwidth, and computational resources. We decompose the mixed-integer nonlinear programming (MINLP) problem into two subproblems, devising a sensing task allocation algorithm and a proximal policy optimization (PPO)-based task offloading and resource allocation (PTORA) algorithm to address them. Comprehensive simulations validate the effectiveness of the proposed PTORA in optimizing information value, reducing task execution delay, and minimizing energy consumption.

Wireless Traffic Prediction (WTP) plays a significant role in achieving intelligent resource management for communication systems. However, WTP still faces challenges such as inaccurate prediction resulting from the complex spatial-temporal characteristics due to user mobility, high communication overhead caused by the complexity of the prediction model, and user privacy issues stemming from Centralized Learning (CL). To address the aforementioned issues, this paper proposes a WTP framework under the Federated Learning (FL) strategy called Federated Spatial-Temporal Dual-attention based Network (FedSTDN). Aiming at improving communication efficiency and simultaneously representing various wireless traffic patterns, a data augmentation-based clustering algorithm is adopted, which groups cells into different regions using a small augmented dataset, facilitating subsequent processing. To improve prediction performance, a local prediction model based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) is proposed to capture the short- and long-term dependencies of traffic. Additionally, a novel Kolmogorov-Arnold Network (KAN) layer is introduced to replace the traditional Multi-Layer Perceptron (MLP) layer, further enhancing prediction performance. Simulations on two different real-world datasets verify the effectiveness and efficiency of FedSTDN. Compared to the well-performing baseline, the proposed FedSTDN achieves up to 32.83% and 24.30% improvements in Mean Square Error (MSE) and Mean Absolute Error (MAE) on the Milan dataset, respectively. For the Trentino dataset, FedSTDN achieves up to 17.25% and 5.86% improvements in MSE and MAE, respectively.

Unmanned aerial vehicle (UAV) relay networks with flexible and controllable characteristics are expected to complement the capacity of the gNB. This paper studies the multiplexing of enhanced Mobile BroadBand (eMBB) and Ultra-Reliable Low-Latency Communications (URLLC) in a multi-UAV relay network, where the strict latency requirement of URLLC can be achieved by the preemptive multiplexing of eMBB resources. However, this may affect eMBB reliability due to the transmission interruptions. Moreover, given the limited energy resources of UAVs, there is an inherent tradeoff among reliability, delay, spectral efficiency, and energy efficiency. To address these challenges, this paper develops a hierarchical UAV-assisted eMBB/URLLC multiplexing scheduling framework. For the eMBB scheduler, we first utilize multiple UAVs to assist the gNB in relaying eMBB traffic and formulate the eMBB resource allocation problem as an optimization problem. Then, we propose a decomposition-relaxation-optimization algorithm to maximize eMBB data rates while considering the personalized fairness of resource allocation and UAV power consumption. For the URLLC scheduler, we further consider the multiplexing of eMBB/URLLC traffic based on the optimization of eMBB resources. To reduce the performance fluctuations of eMBB, we propose a novel cross-slot strategy to schedule URLLC within two time slots rather than one time slot as in existing works. With this strategy, a deep reinforcement learning-based algorithm is proposed to obtain the optimal strategy for the preemption of URLLC on eMBB. Simulation results show that the proposed algorithms outperform the benchmark schemes in terms of convergence rate, eMBB reliability, personalized resource fairness, UAV consumption, and URLLC satisfaction.

In vehicular edge computing networks, the real-time, on-demand scheduling of scarce network resources for environmental perception, task offloading, computation, and feedback is vital. However, these coupled processes make resource allocation challenging. Moreover, existing real-time channel measurement techniques in complex vehicular topologies present load, accuracy, and customization difficulties. To address these issues, this paper proposes an on-demand environmental perception and resource allocation strategy. Specifically, with the introduction of a channel knowledge base, we first analyze the coupling relationship between environmental perception, communication, and computation. A model is then proposed for task offloading to schedule the granularity of environment perception, communication resources, and computational resources dynamically. Subsequently, the resource allocation problem is formulated as an optimization problem, aiming to minimize system processing delay and maximize resource utilization while ensuring perception accuracy. To address this, a two-phase optimization-assisted deep reinforcement learning (DRL) algorithm is proposed. The initial phase uses convex optimization to approximate a solution. The second phase proposes a DRL-based algorithm to intelligently schedule dynamic network resources, with the first phase’s solution guiding the initial exploration space to enhance DRL training efficiency. Extensive simulation experiments verify the effectiveness of our proposal.

Preemptive scheduling efficiently addresses the coexistence of enhanced Mobile Broad Band (eMBB) and Ultra-Reliable Low-Latency Communications (URLLC). While URLLC puncturing influences eMBB performance, further investigation is necessary to study the trade-offs between stability, delay, and efficiency. However, existing studies overlook the imbalance in eMBB/URLLC load distribution and personalized fluctuations in eMBB performance, leading to sub-optimal results. To tackle this, we propose an unmanned aerial vehicle (UAV) relay-assisted eMBB/URLLC multiplexing framework. Specifically, considering the utilization of UAVs for connecting separated next-generation Node Bs (gNBs) and the individual subject experience of services, we first formulate the multiplexing problem as an optimization problem. The objective is to maximize eMBB throughput and minimize personalized fluctuations in eMBB performance and UAV consumption, subject to URLLC constraints. Then, the challenging problem is decomposed into the eMBB problem and the URLLC problem. For the former, we further decompose it into three sub-problems and solve them using optimization methods. For the latter, we propose a deep reinforcement learning-based algorithm to obtain an optimal strategy for relaying and puncturing URLLC into eMBB intelligently. Simulation results demonstrate that our proposals outperform benchmark schemes regarding eMBB throughput, UAV consumption, eMBB performance fluctuation, URLLC satisfaction, and learning efficiency.

Urban air mobility (UAM) provides a new solution to relieve urban transportation pressure by expanding transportation resources of near-ground space. The vigorous development of emerging technologies such as artificial intelligence, intelligent transportation, and sixth-generation (6G) communication technologies have greatly promoted the progress of UAM. However, UAM also increases traffic safety hazards while introducing vertical dimension transportation resources. Traditional collision avoidance is not suitable for three-dimensional (3-D) air-ground integrated mobility scenario, which considers safety hazards in vertical dimensions as well as the resource supply and demand conflict due to the combined effect of directional antenna angle and limited communication distance. Therefore, a safety-oriented on-demand resource allocation strategy for air-ground integrated mobility is proposed. Specifically, we first model the 3-D safety distance model in the air-ground integrated mobility scenario and construct its quantitative relationship with communication and computing resources. Secondly, a 3-D safety distance optimization model is proposed with joint consideration of safety-oriented resource requirements and resource distribution, which can allocate resources in the scenario. Furthermore, a 3-D safety distance optimization algorithm based on deep reinforcement learning (DRL) is designed for solving the optimization model, which implements a safety-oriented resource allocation. Simulation results show that the proposed safety control strategy can effectively improve the safety of air-ground integrated mobility and alleviate the contradiction between the supply and demand of resources.

In digital twins enabled space-air-ground integrated networks (DT-SAGINs), the DT of a vehicle (DT-V) needs to constantly migrate between the infrastructures deployed on the path of the vehicle as the vehicle moves to provide stable and continuous driving services for the vehicle. However, each DT-V has differentiated migration requirements and the heterogeneous network infrastructures have various migration performances. Therefore, how to design a scheme that jointly considers the above factors to determine the optimal migration strategy for each DT-V becomes a challenge. In this paper, we propose a game-based migration scheme for the DT-Vs in DT-SAGINs. In this scheme, we first design a two-layer DT migration architecture, where each vehicle has two DTs and each network infrastructure only has one DT. The two DTs of the vehicle are respectively deployed in the cloud layer (Primary DT-V) and the edge layer (Second DT-V). In contrast, the DT of each network infrastructure is deployed in the cloud layer (DT-I). Based on the designed architecture, the interaction of the Primary DT-Vs and the DT-Is deployed in the cloud layer is formulated as a matching game, where an integrated algorithm that couples bilateral matching and dynamic programming is designed to obtain the optimal migration strategy for each Second DT-V deployed in the edge layer to maximize its average utility. The simulation results show that the proposed scheme can lead to a higher utility for each Second DT-V than the conventional schemes.

The fifth-generation network (5G) has made great progress. With the continuous development of communication technology, by analyzing the characteristics of 5G scenarios, the sixth-generation network (6G) technology combined with multiple scenarios provides effective solutions for the implementation of emerging services with stringent requirements. It is worth noting that the vigorous development of emerging services has been weakened due to the limited resources provided by a single scenario, cross-scenario technologies are urgently needed to enable emerging services in the 6G stage. However, most of the existing work only focuses on a single scenario, which leads to emerging services with complex requirements still difficult to achieve in practice. Therefore, we propose an efficient representation and scheduling framework to achieve the unification of cross-scenario resources, aiming to solve the problem of resource scheduling in cross-scenario. In the above framework, first of all, considering the strict resource requirements of emerging services, we establish a unified resource representation model based on the Time-Expanded Graph (TEG). Secondly, to maximize resource utilization, based on the representation model, a cross-scenario resource scheduling model is proposed. Then, considering the complexity of solving the scheduling model, a resource utilization maximization strategy is presented through the primal decomposition. Simulation results show that the unified framework can effectively improve resource allocation efficiency in complex 6G scenarios.

Abstract MRI and CT images have been routinely used in clinical practice for treatment planning of the head‐and‐neck (HAN) radiotherapy. Delineating organs‐at‐risk (OAR) is an essential step in radiotherapy, however, it is time‐consuming and prone to inter‐observer variation. The existing automatic segmentation approaches are either limited by image registration or lack of global spatial awareness, thus under‐performed when dealing with segmentation of complex anatomies. Herein, we propose a full‐scale attention network (FSANet) that integrates bi‐side skip connections, full‐scale feature fusion modules (FFM), a feature pyramid fusion and supervision module (FPFSM) to accurately and efficiently delineate OARs in HAN region on CT and MRI scans. Specifically, bi‐side skip connections were adopted to keep small targets in the deep network and to capture semantic features at different scales. The FFM with cascaded attention mechanisms were used to recalibrate the significant channels and salient regions in the feature maps. The FPFSM was used to guide the network to learn the hierarchical representation so as to improve the segmentation robustness. The proposed algorithm was validated on the public benchmark HAN CT dataset and an in‐house MR dataset. Both results show significant improvement compared to state‐of‐the‐art OAR single‐stage segmentation methods for the HAN region.

In 5G/B5G networks, the preemptive scheduling provides an efficient solution to the coexistence problem of eMBB/URLLC services. Current works usually assume that the downlink transmission duration of each URLLC service is within one mini-slot, which ignores the different requirements of URLLC users and may lead to the severe data rate loss of eMBB services and low resource utilization efficiency. To deal with above problem, we propose a novel URLLC preemptive strategy, where the arriving URLLC services could cross through multiple mini-slots rather than only one to puncture resources on demand. With the proposed strategy, considering the heterogeneous delay requirements of URLLC services and the preemptive influence on eMBB services, an efficient algorithm based on optimized sparrow search is also proposed. Through allocating time and frequency resources occupied by each URLLC service on de-mand, the number of URLLC services supported by the gNB is maximized while the satisfaction of eMBB services is ensured. The simulation results indicate that the proposed algorithm can achieve better performance compared with the benchmark schemes.