Recipient organisationUniversity of ExeterSource-published name: University of Exeter
Funding£192K
PeriodFeb 2025 — Feb 2027
In plain English
AI plain-English summary
Cars streaming high-definition maps and 360-degree video will learn to predict what content a driver needs next, before the driver requests it. This project tackles a fundamental bottleneck in the Internet-of-Vehicles: vehicles move fast, connections drop frequently, and content requests shift unpredictably. Current systems cannot juggle the competing demands of storing data (caching), processing it (computation), and transmitting it (communications) in real time. The researchers will combine vehicular communication protocols with machine learning to solve this. They plan three technical advances: a unified model to assess and allocate caching, computation, and communications resources; a privacy-preserving federated learning method that analyses vehicle trajectories and service requests without exposing individual driver data; and a deep-reinforcement-learning system that adaptively selects transmission bitrate and video resolution while optimising resource use. If successful, the work could directly improve real-time navigation, high-definition mapping, and auxiliary-driving video for everyday drivers. It would also provide automakers and mobile network operators with standardised technical frameworks for deploying these systems at scale. The project is applied engineering—it aims to produce market-ready solutions, not fundamental science.
View original technical description
Real-time multimedia service optimization in the Internet-of-Vehicles (IoV) has attracted significant interest from both academia and industry due to the constantly increasing demands on vehicular multimedia applications such as real-time navigation, high-definition maps, and 360-degree video for auxiliary driving. The future IoV is envisioned to incorporate caching, computation, and communications (3C) resources to handle real-time multimedia transmission effectively and to create a safe, effective, and comfortable driving environment. However, the prediction, caching and delivery of streaming media contents confront significant obstacles, given the high mobility of vehicles, intermittent information transmission, high dynamics of content requests, and complexity of working scenarios. To address these challenges, this project aims to seamlessly synergize the benefits of vehicular communication protocols and Machine Learning technology in order to maximize the 3C resource utilization, enhance the Quality-of-Service, and protect user privacy. First, a robust 3C resource integration mechanism empowered by edge intelligence with a unified assessment model will be proposed to optimize real-time vehicular multimedia services. Second, a privacy-preserving federated multi-modal learning approach will be developed to analyze the spatial-temporal connection between vehicle trajectory and service requests and create a reliable traffic prediction model. Third, a hierarchical Deep Reinforcement Learning based decision-making method will be developed to construct an adaptive multimedia transmission control model that can simultaneously optimize the 3C resource allocation as well as the selection of transmission bitrate and reconstruction resolution. The outcomes of this project will produce cutting-edge theoretical and technical advancements that will aid in technical advancement, standardization, and market application of automakers and mobile communication service providers.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know