In developing Advanced Driver Assistance Systems (ADAS) and Automated Driving Systems (ADS), traffic sign detection is a critical task. However, existing methods face limitations in real-time performance and small-object detection accuracy, partly due to deployment platform constraints. Our proposed Neck-Heavy YOLOv5 (NH-YOLOv5) introduces significant improvements. It features a hybrid data augmentation approach that enhances the detection of small objects. Moreover, a unique neck-heavy network architecture is designed to optimize detection performance without increasing computational complexity. Notably, NH-YOLOv5 is integrated into the digital twin framework of traffic systems. By leveraging the digital twin’s ability to simulate and monitor traffic scenarios in real time, the model can continuously adapt and improve its detection capabilities. Experimental results based on the Tsinghua - Tencent 100K (TT100K) dataset demonstrate that our method achieves a 53% reduction in computation while enhancing accuracy and small object detection ability, making it highly suitable for real-time traffic sign detection in the context of digital twins.

The Internet of Vehicles (IoV) presents a rapidly evolving ecosystem that enables seamless communication among vehicles, infrastructure, and pedestrians. However, the open nature of the IoV exposes it to various security threats, including the spread of malicious data and trust violations. To address these challenges, we propose a Filter-Based Trust Evaluation (FTE) system, which combines direct and indirect trust evaluation metrics to assess the trustworthiness of vehicles. The proposed system utilizes a Kalman filter-based approach to fuse predicted and measured trust values, allowing for dynamic and accurate trust assessments. Our extensive simulations show that FTE outperforms existing trust management schemes, demonstrating superior accuracy, efficiency, and robustness, even in environments with high malicious vehicle ratios. The results highlight the potential of FTE for secure and reliable operation in the IoV systems.

In disaster scenarios, secure and reliable data collection in Vehicular Ad Hoc Network (VANET) is crucial, yet the network often suffers from issues such as infrastructure damage, network partitioning, and vulnerabilities to attacks (e.g., False Data Injection and Black Hole Attack). Trust management is a promising solution to prevent these attacks. However, in infrastructure-less and partitioned disaster areas, existing trust schemes face problems of trust evidence sparsity and evaluation inconsistency, leading to inaccurate detection. To address these limitations, we propose an Unmanned Aerial Vehicle (UAV)-Assisted Traceable and Hierarchical Trust Management scheme (UATHTM). The UATHTM includes a Vehicle-to-Vehicle (V2V) local trust model and a UAV-to-Vehicle (U2V) global trust model. The former facilitates rapid detection of false data, while the latter is designed for accurately tracing malicious vehicles. Specifically, the V2V model constructs a mutual adjustment between entity-centric and data-centric assessments, continuously refining trust for accurate local detection. The U2V model incorporates a new trust metric based on disaster trajectory similarity to enhance the accuracy of global tracing, through leveraging the comprehensive view of UAVs. Extensive simulations demonstrate that the UATHTM scheme outperforms existing trust management schemes, showing higher precision, recall, and F1-score in detecting false data and malicious vehicles in VANET under challenging conditions.

Over the past few decades, as vehicles have become increasingly intelligent, the applications of in-vehicle electronic systems have expanded significantly. However, with the growing complexity of vehicle networks, there is an ever-increasing concern for their network security. In particular, the Controller Area Network (CAN) bus has become a critical medium for communication between various Electronic Control Units (ECUs) within a vehicle. Since the design of the CAN bus lacks sufficient security measures, it is vulnerable to various network intrusions. To address this security challenge, researchers have been searching for ways to enhance the network security of the CAN bus to ensure that vehicle systems are not compromised by unauthorized access or network attacks. This paper introduces a robust intrusion detection system (IDS) for the CAN bus in vehicles, employing a novel Multi-Scale Feature Fusion technique. Leveraging the distinct capabilities of Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Transformer neural network architectures, the proposed methodology adeptly captures and prioritizes both shallow and deep features of CAN bus data. Besides, it enhances detection accuracy and robustness against various cyber-attacks. Evaluation on the Carhacking dataset demonstrates superior performance, achieving a precision, recall, and F1-score of 100%. Verified by an ablation study, this approach promises a substantial advancement in safeguarding in-vehicle networks.

With the development of driverless technology, unmanned ground vehicles (UGVs) and unmanned aerial vehicles (UAVs) have been widely used in the realm of commodity delivery. In the scenario with a high requirement for timeliness commodity delivery, the mobile unmanned retail mode has grabbed tremendous sights. However, the traditional retail mode solely focuses on presetting routes for delivery, but fails to deal with the real-time change of customers orders with the assistance of UAVs and UGVs. In this paper, we propose an air-ground cooperative real-time delivery scheme based on joint scheduling. Specifically, UGVs deliver commodities to customers based on spatial-temporal costs (e.g., delivery distance and order urgency). The UAV serves to replenish UGVs with commodity resources, achieving a balance between regional resource consumption and UAV resource replenishment based on joint scheduling. Finally, experimental results show our scheme outperforms other baseline schemes in terms of customers average waiting time, UGVs average delivery delay time, and UAV total flight distance.

In recent years, Electric Vehicles (EVs) have emerged as a sustainable alternative to internal combustion vehicles, noted for better efficiency, lower operational costs, and reduced carbon emissions. However, with the growing adoption of EVs and limited charging infrastructure, challenges such as charging congestion arise. Traditional plug-in and in-Parking Vehicle-to-Vehicle (V2V) charging modes, constrained by fixed charging locations, lack flexibility and necessitate long charging times. Therefore, this paper introduces a novel in-Motion V2V charging mode, termed V2V (M) mode, allowing an EV as an energy Provider (EV-P) and an EV as an energy Consumer (EV-C) to form a V2V charging Pair (V2V-Pair). Then, the V2V-Pair can transfer energy via wireless V2V charging service while on-the-move. In this paper, the proposed V2V (M) management framework employs a Path Proximity-based V2V Pair matching algorithm and spatio-temporal cooperative path planning, to enhance charging efficiency and reduce charging trip duration. The urban environment simulation results demonstrate marked improvements of the proposed V2V (M) mode. It shorters the charging trip duration and enhances charging service efficiency, offering a viable solution to current EV charging constraints.

The Internet of Vehicles (IoV) has witnessed a significant growth in the number of participants. This rapid expansion has increased demands for computing resources and quality of service (QoS), posing challenges for mobile edge computing (MEC) in the IoV domain. Efficiently allocating computing power to meet these service demands has become a crucial concern. Therefore, joint optimization of offloading decisions and power allocation is required to achieve the tradeoff between task latency and energy consumption. To address the above challenge, we propose a multi-agent reinforcement learning (MARL) method called multi-agent twin delayed deep deterministic policy gradient (MA-TD3) in this paper. Compared to its predecessor, multi-agent deep deterministic policy gradient (MADDPG), this algorithm improves performance and execution speed. It solves the slow convergence problem caused by Q-value overestimation and reduces the computational cost. The experimental results illustrate that the proposed algorithm reaches an observable performance improvement.

There is an increasing need for charging station recommendation to minimize the overall charging time for electric vehicles and balance load for the charging stations. To grant this need, we model the recommendation problem as a Markov Decision Process (MDP) problem. However, the traditional MDP model has the issue of ‘curse of dimensionality’. To address this issue, we propose an extension of MDP: multiple-phase MDP, in which the state transition of MDP is decomposing into several phases, so as to reduce the state space and state transition complexities. This is done by introducing two states other than the normal state defined in MDP: post decision state and intermediate decision state. Then, we propose an online learning based algorithm to solve the formulated multiple-phase MDP model. Thanks to the reduced complexities of the state space and state transition, the proposed online algorithm can converge fast. By comparing to other recommendation mechanisms, such as game theory based recommendation and Q-learning based recommendation, our simulation evaluation demonstrates that our proposition can bring good performance.

User mobility management is one of the important components of mobile multimedia systems. In an IP-based mobility case, a mobile node should be able to seamlessly obtain reserved resources after handover to a new access router. This is essential for both service continuity and quality of service assurance. In this paper, we introduce a hybrid architecture which allows a mobile node to seamlessly obtain reserved resources from its new location (seamless reservation), as well as to quickly release obsolete resources along the old path. In this architecture, reservation on access networks is performed by using path-decoupled approach, while path-coupled approach is used on backbone networks. Moreover, instead of using sender-initiated or receiver-initiated approach, mobile-node-initiated approach is used in this architecture to further improve reservation performance. In our performance analysis, we analyze advance reservation delay and release delay, as well as reservation underutilization.

Reservation aggregation provides scalability to the IETF integrated services (IntServ) by reducing the high number of states stored at internal routers and the number of signalling messages processed at these routers. However, the latter benefit will be lost if the bandwidth of aggregate reservation changes frequently. Hence, previous works either hold resource requests during a waiting period before sending a single aggregate reservation for all received requests, or reserve maximum resources which will be requested in the following period. However these works involve an assumption of arrival distribution of resource requests. In this paper, we design a policy for enhancing the reservation aggregation performance, which is independent of arrival distribution. From the results of simulation, we observe that this policy outperforms other policies.

Resource reservation in mobile environments is an important task in the future when the user’s point of attachment to the network changes frequently due to mobility. One crucial issue of this task is to release unnecessary reservation along old path after departure of mobile node(s). Although this can be done automatically by a soft state mechanism, default soft state’s lifetime which is conceived for wired network case is not suitable to mobility. In this paper, we introduce an additional value to soft state’s lifetime to optimize resource release along old path: besides the default value which is used when no handover occurs, the additional value is assigned when handover takes place. Through the comparison with three other mechanisms, we observe that our proposed mechanism outperforms the others in term of average cost and blocking probability.