In this letter, we address efficient communication and resource allocation for vehicular networks in urban environments. Due to spectrum reuse and rapidly varying interference, joint spectrum allocation and power control become particularly challenging in such scenarios. To tackle this problem, we formulate a system-level optimization framework and propose a multi-agent reinforcement learning approach termed the Predictive Residual Q-Network Method (P-RQNM). An LSTM-based predictor is employed to predict short-term interference from historical observations, while a residual Q-network refines individual Q-values using trajectory statistics to improve decision consistency and contribute to more stable training behavior. Simulation results under urban vehicular scenarios show that P-RQNM outperforms baseline multi-agent value-decomposition methods in terms of link capacity, transmission reliability, and convergence behavior, demonstrating robust performance across diverse traffic densities and loads and practical potential for urban vehicular networks.

In multi-access edge computing (MEC) environments, optimizing task offloading and communication-aware resource allocation is challenging due to dynamic topologies, fluctuating channels, and heterogeneous link capacities. This letter proposes a Graph-Augmented Multi-Agent Proximal Policy Optimization (GA-MAPPO) framework under a centralized training and decentralized execution (CTDE) paradigm. A Master Agent cooperative architecture integrates a Graph Attention Network (GAT) to model dynamic communication dependencies and bandwidth variations among edge nodes. The hybrid action space enables joint optimization of task offloading and transmission resource allocation. Experiments based on mobility and workload traces show that GA-MAPPO improves task completion rate and reduces latency compared with representative baselines, demonstrating its adaptability in dynamic MEC and Internet of Vehicles (IoV) scenarios.

In recent years, the demand for smart healthcare solutions have heightened the need for accuracy, reliability, and comfort in bedside ECG recording and analysis. This study presents a bedside non-direct contact ECG recording system based on capacitive coupling electrocardiography (cECG) and verifies its performance in accurately capturing Heart Rate Variability (HRV) during the night. Firstly, cECG collects ECG data through clothing, avoiding skin irritation from conventional wet electrodes. Secondly, leveraging the unique characteristics of cECG signals, a deep learning framework assesses the quality of cECG, filtering noise and identifying off-bed information, enhancing HRV analysis precision. Subsequently, the system was employed to recording sleep data from 6 subjects overnight, with our proposed algorithm utilized for signal quality assessment (SQA) and HRV analysis. Finally, HRV features were compared with synchronously collected wet electrode ECG signals, encompassing time domain features, frequency domain features, and nonlinear features, totaling 13 HRV features. Experimental findings demonstrate that for the SQA task, the model achieved a classification accuracy of 94.7%, with a Recall of 0.941, Precision of 0.940, F1 score of 0.941, and Cohen's Kappa of 0.927. The accuracy of on/off-bed monitoring reached 99.79%. Additionally, HRV features showed a strong correlation with the reference ECG. In the time-domain metrics, the largest mean absolute percentage error (MAPE) is for PNN50, with a value of 8.148%. In the frequency-domain features, the largest MAPE is for HF, with a value of 13.253%. For nonlinear features, the largest MAPE is for SD1, with a value of 5.182%. Generally, the system exhibited a reliable solution for cECG recording, on/off-bed status detection, and bedside HRV analysis.

Non-orthogonal multiple access (NOMA) is a promising technology garnering significant attention among the Internet of Things (IoT) community due to its superior spectral efficiency. This work addresses the rate region boundary enhancement of downlink NOMA systems under imperfect successive interference cancellation (SIC) with advanced improper Gaussian signaling (IGS), which provides additional degrees of freedom for system design. We investigate a universal scenario in which two users adopt improper signaling and their transmit powers are optimized. We first formulate the achievable rate of both users in terms of the impropriety degree of the IGS. First, the analytical expressions for the best improper transmission are characterized by jointly optimizing the users’ power and the impropriety degree for the perfect SIC case. Then, a deep Q network (DQN)-based approach is provided to find the rate region of the IGS-aided NOMA system under imperfect SIC. Simulations presented for the downlink NOMA system support the analysis, illustrating that IGS can efficiently enhance the rate region of the NOMA system compared to proper signaling.

It is a critical issue for achieving dynamic average consensus (DAC) in the presence of privacy eavesdroppers and false data injection (FDI) attacks, and this scenario is applicable to intelligent transportation systems. A dynamic event-triggered privacy preserving DAC (DET-PPDAC) control scheme is proposed. Firstly, in a privacy-sensitive scene, different time-varying terms are added to communication states by hiding real information from eavesdroppers. An observer and a compensator are designed to construct a control scheme for compensating for the impact of FDI attacks over a channel between a control signal and an actuator. Adaptive auxiliary variables are introduced for compensating for residual errors owing to asymmetric encryption/decryption functions. Dynamic event-triggered conditions are constructed to reduce the number of data as well as the risk of leakage by eavesdroppers, and continuous monitor from neighbors is removed. Our DET-PPDAC control scheme can also be applied to a directed graph. Stability analysis shows that the control scheme finally achieves DAC with bounded errors and Zeno-free behaviors while satisfying the requirement of privacy preservation. Simulation examples with formation of vehicles are given to demonstrate the effectiveness of the proposed control scheme.

为了保障智能网联汽车的低时延通信,利用蜂窝车联网中V2X(vehicle to everything)信道模型、边缘计算技术,研究计算卸载与边缘缓存联合优化策略.设计了一种智能网联汽车计算卸载与边缘缓存协同模型L-DDPG(least-deep deterministic policy gradient),通过对车载本地与边缘计算资源的整合,支持V2X场景下对不同计算任务的分类处理.由边缘平台对车载计算请求进行预判决,确保对连续的计算任务快速响应;结合基于最近最少使用(least recently used LRU)的边缘缓存策略,实现对新计算任务的高效管理;基于DDPG算法对计算卸载与边缘缓存进行联合卸载决策.仿真结果表明:L-DDPG模型性能优于传统模型,能够有效提升系统的工作效能,在保障业务服务质量的同时降低时延和系统资源消耗.

随着城市化进程不断加快,城市功能分区正从规模增长向质量提升转型发展,优化城市空间布局、深化多模式融合交通运营和打造一体化出行服务平台是城市交通数字化转型的核心需求,挖掘城市交通出行特征并对其进行分析,有助于完善城市立体化交通服务体系、满足多样化出行需求、推进城市用地合理开发利用以及指导城市决策者制定合理的规划措施.蜂窝信令数据(Cellu-lar Signaling Data,CSD)具有覆盖范围广、样本量大、可长期连续监测等优势,蜂窝网络大数据可以较低的成本挖掘大规模人口的起讫(Origin Destination,OD)分布和出行行为模式,是促进未来城市智能交通发展的重要组成部分.基于此,对现有交通信息采集方法、发展制约因素和蜂窝信令数据价值进行了总体概述,梳理了面向蜂窝网络大数据的城市交通出行特征挖掘框架、关键技术研究进展和未来发展方向.首先,依据基于蜂窝信令数据的城市交通出行特征挖掘系统的功能规划与发展需求给出了其架构设计与应用框架.其次,从蜂窝网络出行链构建角度总结了蜂窝移动通信网络结构、出行链特征与提取框架,阐述了针对出行链中噪声数据和轨迹震荡的数据优化方法,以及出行链轨迹与实际路网融合时的路网匹配技术.然后,面对蜂窝网络大数据驱动下的城市空间结构优化与多模式交通发展需求,从人口流动监测、出行模式识别、行为分析与预测等城市交通出行特征挖掘方面详细介绍研究现状.最后,从5G优化定位、多源数据处理与挖掘、细粒度出行模式识别、基于组件的系统模型体系构建等方面指出了未来研究的技术方向和发展趋势.

The simultaneous wireless information and power transfer (SWIPT) relay system is one of the emerging technologies. Xiaomi Corporation and Motorola Inc. recently launched indoor wireless power transfer equipment is one of the most promising applications. To tap the potential of the system, hybrid automatic repeat request (HARQ) is introduced into the SWIPT relay system. Firstly, the time slot structure of HARQ scheme based on full duplex two-way amplify and forward (AF) SWIPT relay is given, and its retransmission status is analyzed. Secondly, the equivalent signal-to-noise ratio and outage probability of various states are calculated by approximate simplification. Thirdly, the energy harvesting power in each state is calculated. Finally, the energy harvested-throughput sum function is constructed to characterize the performance of energy harvesting and data transmission. Simulation results show that the proposed HARQ scheme has better energy harvested-throughput sum function than the traditional HARQ scheme. When P2 = 22 dB, the maximum sum function is 54.86% (the proposed HARQ scheme) and 52.307% (the traditional HARQ scheme), respectively.

针对智能网联车辆高速移动以及智能网联组网模式多元化导致的传统协同过滤算法有效性受到限制的问题,提出一种新型混合标签感知推荐模型(hybrid tag-aware recommender model,HTRM).嵌入层采用Word2Vec模型对项目标签、项目评分、用户行为标签和用户评分进行向量表示;特征层引入自编码器提取项目的自相似特征,采用长短期记忆网络(long short-term memory,LSTM)提取用户行为特征;门控层联合用户和项目的特征,并输入至全连接神经网络(fully connected neural network,FCNN)进行评分预测.实验结果表明,与TCF、CCF、ACF和DSPR传统模型相比,HTRM模型设计更合理,可以获得较高的推荐预测精度.

As an important part of urban intelligent transportation system, predicting users' demand for taxi using offline GPS data has attracted interests in recent years. In general, the distribution of traffic flow in different areas of the urban is different. Furthermore, the characteristics of traffic flow in different location areas present different. Therefore, it is challenging to achieve accurate prediction of users' demand for taxi in different spatiotemporal scenes. This paper presents a combined multi-scale residual calibration network (MS-ResCnet) using residual calibration network and multi-scale fusion mechanism to predict users' demand for taxi. Specifically, mapping rasterization and time series division methods are used to convert vehicular GPS data into spatiotemporal images of traffic flow within continuous sub grid areas. Then, the proximity, periodicity, and tendency, as significant characteristics of spatiotemporal images of traffic flow in each sub grid area are extracted. Thus, we establish a multi-dimensional spatiotemporal characteristic perception scheme of traffic flow in each grid area. Moreover, the datum characteristics and calibration characteristics of spatiotemporal images are extracted through the dual channel ResCnet network. Through the deep stagger training network, the full fusion of multi-scale spatiotemporal characteristics is realized. The performance of MS-ResCnet model is evaluated and verified using public datasets. The simulation results show that the traffic flow prediction performance of MS-ResCnet model is better than that of traditional STAR model. The root mean square error (RMSE) of the proposed method outperform the STAR model around 2.57%.

为改善蜂窝车联网(C-V2X)频谱利用效率,提出了一种针对系统下行吞吐量最大化、并保证车对车通讯(V2V)链路连接性的资源分配算法.定义蜂窝车联网信道模型,在最大发射功率、中断概率等约束条件下建立优化模型并对其进行分步求解,采用二分图最佳匹配(KM)算法动态调度信道资源,实现C-V2X通信系统中车辆到路边设施(V2I)下行链路与V2V链路之间动态分配网络资源,最后通过不同交通场景评估算法的优化性能.结果表明:本算法经7次迭代后进入稳态,在保证V2V链路连接性条件下实现了V2I链路资源的优化分配,下行链路频谱效率相较于贪心算法平均提升3.5%以上.

The driver's emotional state directly affects safe driving. Under the "vehicle-human-road-cloud" integrated control framework, we propose an end-edge-cloud collaborative emotion perception network model (EEC-Net). The end side extracts the key frames of the driver's face video stream and performs batch compression; the edge side extracts the region of interest (ROI) of the reconstructed images as the input of the emotion recognition model (tiny_Xception) for classification; the cloud control terminal receives abnormal ROI image data and performs online training to dynamically adjust the operating parameters of the edge model. Finally, we test on open and self-built datasets, and the results show tiny_Xception has a significant improvement in accuracy of 2.45% compared to mini_Xception; the EEC-Net model can reliably perceive the negative emotion period, and the overall system memory consumption is reduced by about 5%, and the network transmission data volume and the computation time of emotion recognition are reduced by 95%, 60% respectively.

With the rapid growth of users and sustained network demands powered by different industries, the quality of service (QoS) of the cellular network is affected by network traffic and computing loads. The current solutions of QoS improvement in academia focus on the fundamental algorithms within the physical and medium access control (MAC) layer. However, traffic features of various scenarios extracted from field data are rarely addressed for practical network configuration refinement. In this paper, we identify significant indicators of high traffic load cells according to the field data provided by telecommunication operators. Then, we propose the analysis flow of high traffic load cells with basic principles of network configuration refinement for QoS improvement. To demonstrate the proposed analysis flow and the refinement principles, we consider three typical scenarios of high traffic load cells, including high population density, emergency, and high-speed mobility. For each scenario, we discuss traffic features with field data. The corresponding performance evaluation demonstrates that the proposed principle can significantly enhance the network performance and user experience in terms of access success rate, downlink data rate, and number of high traffic load cells.

The integrated satellite-terrestrial relay network (ISTRN) is a necessary part of the next-generation wireless communication system, and has important practical significance for accelerating the construction of my country's air-space-terrestrial integrated network system.In the traditional ISTRN architecture, a large amount of signaling needs to be forwarded to the ground control center for processing, which increases the delay of network control and management.A new cloud fog computing architecture was proposed, which constructs a sub-regional edge fog computing layer between the ground access and the central cloud to improve the flexibility of business flow management and control.Under the cloud network framework, a Q-learning based edge computing offloading strategy was designed, and the offloading performance was evaluated by time delay and energy consumption.Simulation results show that, compared with Min-min algorithm and backtracking algorithm, Q-learning based computational offload algorithm has better performance in terms of time delay and energy consumption, and can achieve a balance between the joint optimization of time delay and energy consumption.

随着汽车产业电动化、智能化、网联化、共享化的发展驱动,全球主要强国均将智能网联汽车列为国家战略发展方向.蜂窝车联网、边缘计算网络和高精度定位系统的技术发展,为车车、车路、车人和车云系统的全面融合提供了有效支撑.车辆、道路、云平台与蜂窝车联网（Cellular vehicle-to-everything, C-V2X）网络的融合,加速打通车内与车外、路面与路侧、云上与云间的信息互通,为实现车路云一体化的融合感知、群体决策及协同控制提供了重要基础.首先,梳理了智能网联车路云协同系统架构与关键技术,对该领域的演进特征、发展制约因素进行了总体概述;其次,阐述了新型车路云协同系统、智能网联C-V2X通信系统、云控系统和车路云协同测试系统的架构设计与工作原理;然后,从C-V2X组网、融合定位、测试评价角度,介绍了车路云协同系统融合V2X网络、融合定位的技术演进与研究进展,给出了智能网联场景的仿真平台、实车测试及评价指标;最后,对智能网联车路云协同系统的协同组网与控制、互操作、边缘智能服务和安全技术层面的发展趋势进行了展望.

为改善道路交通监测和保证智能网联交通系统的安全、可靠与稳定运行,提出了在路侧边缘平台中基于多通路高分辨率网络与注意力机制融合的车辆检测模型MCHRANet.该模型采用多通路的高分辨率网络的结构设计,保留高分辨率特征并保障识别准确率.融入注意力机制的特征融合方法,通过特征连接权重自学习实现多尺度特征的深度融合.各通路网络采用跳跃连接促进跨层特征融合,加速网络收敛,并利用公开数据集对车辆检测性能进行评估并验证.结果表明:所提模型的车辆检测性能优于3个传统模型,改进后的网络识别平均精度均值(mAP)指标接近95％,且对于不同场景下的检测具有良好的鲁棒性.

In order to solve the problem of unmanned aerial Vehicle (UAV) cluster saturation attack, the autonomous maneuver strategy of UAV is studied. Based on the distributed partial observable Markov decision process, the autonomous maneuver decision model of UAV is constructed, and the Recurrent Multi-Agent Deep Deterministic Policy Gradient (RMADDPG) is used to learn the maneuver strategy of UAV cluster saturation attack. Combined with the typical characteristics of saturation attack scenario, two reward functions of global reward and local reward for UAV maneuver are designed. The trained UAVs can effectively perform suicide saturation attack task. According to the mission target design evaluation index of saturation attack, the Monte Carlo analysis method is used to compare the proposed method with particle swarm optimization (PSO) algorithm in three aspects: inter aircraft collision avoidance, NFZ avoidance and simultaneous attack. The experimental results show that the trained UAVs can effectively perform the suicide saturation attack task under the condition of a small amount of communication, which provides a method for the application of multi-agent deep reinforcement learning to UAV cluster operations.

交通运输环境下的运动车辆检测是近年来计算机视觉以及图像处理领域研究的热点.随着交通联网的普及,交通运输行业对运动目标检测的精确性以及对复杂背景环境的适应性的需求越来越高.因此围绕如何提高交通运输环境下运动目标检测和背景检测的准确度两个方面进行研究,对帧差法和背景差分法改进后融合,并将融合算法运用到复杂的交通环境中.以大量交通录像视频作为样本,用改进的融合算法进行了测试.实验结果表明该算法能够精确快速的检测出目标,完整还原出物体轮廓,有效减小噪声影响,具有良好的抗干扰性.

Tag information is of great significance to improve the performance for rating prediction, but the tag extraction in the recommendation system suffers from heterogeneous information fusion. This paper proposes a hybrid tag-aware recommender model (HTRM). First, the word embedding model embeds the score and the label respectively; then, the autoencoder extracts the text feature of the item label and uses the Long Short-Term Memory (LSTM) to extract the feature of the user label behaviors. The experiments are carried out on the three recognized datasets, namely Last.Fm, Delicious and MovieLens- 20m. Compared with the four algorithms, i.e., TCF, CCF, ACF, and DSPR, the experimental results demonstrate that HTRM significantly outperforms all the baselines.

The SWIPT (simultaneous wireless information and power transfer) DF (decoding and forwarding) relay system could achieve the purpose of both increasing revenue and reducing expenditure. By analysing the system model and transmission characteristics of full-duplex relay, this paper optimizes the retransmission slot structure to enhance the system performance. Firstly, the state transition model is established based on the analysis of the retransmission slot structure. Secondly, the state probability of each state and the transition probability between states are calculated to obtain the total data passing rate, energy transmission efficiency, and total transmission time. Thirdly, in order to compare the performance of various HARQ (hybrid automatic repeat request) schemes more effectively, JNTP (joint normalized throughput of information transmission and energy transmission) is constructed. Monte Carlo simulations finally confirm that the proposed tradeoff-HARQ scheme outperforms the regular-HARQ scheme in terms of JNTP: the performance of the tradeoff-HARQ scheme is 0.03883 higher than that of the regular-HARQ scheme when the total power limit is 20 dB and 0.00651 higher than that of the regular-HARQ scheme when the total power limit is 30 dB.