Shun-Ren Yang

Traffic forecasting is a key task in the field of Intelligent Transportation Systems. Recent research on traffic forecasting has mainly focused on combining graph neural networks (GNNs) with other models. However, GNNs only consider short-range spatial information. In this study, we present a novel model termed LSTAN-GERPE (Lightweight Spatio-Temporal Attention Network with Graph Embedding and Rotational Position Encoding). This model leverages both Temporal and Spatial Attention mechanisms to effectively capture long-range traffic dynamics. Additionally, the optimal frequency for rotational position encoding is determined through a grid search approach in both the spatial and temporal attention mechanisms. This systematic optimization enables the model to effectively capture complex traffic patterns. The model also enhances feature representation by incorporating geographical location maps into the spatio-temporal embeddings. Without extensive feature engineering, the proposed method in this paper achieves advanced accuracy on the real-world traffic forecasting datasets PeMS04 and PeMS08.

Effective driver identification is one of critical aspects of Internet of Vehicles (IoV) applications, playing a pivotal role in various contexts, such as vehicle anti-theft, fleet management, personalized insurance, vehicle settings automation, digital forensics, and so on. In this article, we propose CPBW, a novel mechanism that combines change point detection and Bag-of-Words (BoW). The CPBW utilizes the smartphone triaxial accelerometer data to accurately identify drivers. The key innovation of CPBW lies in its exceptional efficiency within short time windows, significantly enhancing the real-time performance. The study adopts naturalistic driving studies, collecting the unrestricted real-world data to increase applicability. However, challenges arise from dynamic urban environments influencing driving behavior and the need to balance hardware costs, privacy concerns, and data reliability. In comparison to the previous methodologies, CPBW demonstrates a reduced time requirement for driver identification. Particularly, our proposed CPBW mechanism showcases impressive performance, achieving accuracy, precision, recall, and F1-score up to 98.1%, 98.1%, 98.1%, and 98.0%, respectively. As a result, CPBW markedly enhances the practicality of driver identification in real-world scenarios.

As technology advancements are leading to the creation of 5G and next-generation base stations (BS) that offer improved performance and application integration, current solutions are mostly reliant on established technical standards. By incorporating intelligent wireless resource management technology, the current small cell system can be optimized and its transmission performance enhanced. The implementation of deep reinforcement learning was then added. By using indication reports as the state, the smart agent is able to dynamically select the optimal GF parameters to achieve high-efficiency transmission. In the context of ultra-reliable low latency communication (URLLC) applications, we have utilized 5G ns-3 simulation to simulate an IIoT factory scenario that diverges from traditional uplink methods. By implementing grant-free (GF) techniques, we can reduce delays while maintaining a suitable level of reliability. To dynamically select the most appropriate transmission mode under varying conditions, we have developed reinforcement learning (RL) methods. Our numerical results demonstrate a promising trend in the overall satisfaction rate.

Recently, several Internet of Things (IoT) service platforms have been proposed to facilitate IoT application deployment. These platforms typically utilize the lightweight MQTT or CoAP application protocol, optimized for massive IoT applications. Unfortunately, these protocols are not suitable for the emerging, more sophisticated Artificial Intelligence of Things (AIoT). Session initiation protocol (SIP), in contrast, is a signaling and controlling protocol for real-time multimedia sessions, and has been viewed as a better candidate to provide a full range support of different broadband, critical, and industrial AIoT applications. However, there exists no generic SIP-based AIoT service platform that supports creations and operations of heterogeneous AIoT applications with various quality of service. This article presents the first SIP-based AIoT service platform, AIoTtalk, that enables rapid development of scalar and multimedia AIoT applications. Moreover, we deploy an experimental testbed and two real SIP-based AIoT applications to demonstrate the applicability and the performance of our AIoTtalk under both the cloud and edge scenarios. The experimental results show that, together with accurate model predictions and edge-virtualization auto scaling, AIoTtalk guarantees low latency and high quality of experience for messaging and streaming-based AIoT applications.

ETSI Multi-access Edge Computing (MEC) provides an IT service environment and cloud-computing capabilities at the edge of the mobile network, enabling application and content providers to deploy new use cases, like intelligent video acceleration, with low latency and high bandwidth. Specifically, ETSI MEC introduces an MEC server that implements the edge-cloud platform to host partial server-side service logics in the form of MEC applications (MEC Apps). In this paper, we aim to implement the first Proof-of-Concept (PoC) in the literature for the MEC-enhanced mobile video streaming service. Our PoC consists of Android User Apps, an MEC App, and the YouTube server. The MEC App implements two main functions: popular video caching and radio analytics/video quality adaptation. The User App provides general functions of a YouTube video streaming app and can access the videos from the cache server or the YouTube server under the MEC server's guidance. In addition to the PoC implementation, this paper further develops two machine learning models to be incorporated into the MEC App for popular video prediction and radio channel quality prediction, which allows to consider the effect of non-negligible round-trip times and adjust the video quality more accurately. The experiment results justify that our models, together with other advantages from MEC, can guarantee good performance for the mobile video streaming service. Finally, we model and investigate the effectiveness of the MEC architecture for improving the quality of experience (QoE) of video-streaming users.

In the literature, future vehicle trajectories have been exploited to improve the performance of challenging infrastructure-to-vehicle (I2V) group message delivery over vehicular ad-hoc networks (VANETs). The existing trajectory-based I2V group message delivery protocols typically consist of two phases. Based on the reported vehicle trajectories, phase one identifies an appropriate rendezvous point of the group message and each of the member vehicles. Then, phase two composes a multicast tree to deliver the message to these selected rendezvous points. The message is kept at the rendezvous points for later last-hop delivery when the member vehicles pass through. However, separately handling the rendezvous point selection and the multicast tree construction may significantly restrict the overall message delivery performance. This paper designs a novel trajectory-based I2V group message delivery protocol, i.e., eTGMD, for delay-tolerant vehicular applications. eTGMD iteratively selects a promising rendezvous point and extends the current multicast tree to cover the selected rendezvous point. Our specific contributions include the following: 1) We prove the trajectory-based I2V group message delivery problem's NP-hardness, 2) we theoretically analyze the worst-case performance bound and time complexity of eTGMD, and 3) we conduct extensive simulation experiments to evaluate eTGMD's message delivery performance.

IEEE 802.16m resource scheduling remains a challenging issue for video multicast. Given the OFDMA frame structure, the IEEE 802.16m scheduling task is comprised of: (1) intraframe scheduling; and (2) interframe scheduling. In the literature, while many studies concentrated on the development of intraframe scheduling mechanisms, few studies looked at the potential of interframe scheduling mechanisms. This paper is the first attempt to investigate the energy efficiency potential of interframe scheduling algorithms to support scalable-video multicast services over IEEE 802.16m networks. Under the premise that the bandwidth requirements of scalable-video subscribers must be satisfied, we first prove that the interframe scheduling problem of minimizing video-subscriber energy consumption is NP-hard. To tackle the NP-hard problem, we propose a multiple bin-packing algorithm, MBPA, for energy-efficient scheduling. The proposed MBPA has full compatibility with the existing intraframe scheduling mechanisms and with the IEEE 802.16m sleep-mode operations. By applying the divide-and-conquer strategy, MBPA effectively eliminates unnecessary wake-up periods and unnecessary state transitions (between wake-up and sleep states), and thus achieves high energy-efficiency. Through theoretical analysis, we show that MBPA is a p-approximation algorithm, where p is a finite value no less than one. Finally, the simulation results show the effectiveness of the proposed MBPA in energy efficiency, user satisfaction, and computational complexity.