The project has fully achieved its objectives and milestones for the period. In details,
• Task 1 proposed optimal video quality allocation to reduce the required bandwidth of traffic video transmission without noticeable visual quality degradation. Specifically, the fellow first conducted a comprehensive survey on the existing Quality of Experience (QoE) model for video services, which can be classified into subjective test, objective quality model, and data-driven quality model. Then the fellow proposed to allocate dif-ferent qualities to different macroblocks within the traffic monitoring video to reduce the required bandwidth without noticeable visual quality degradation, which conforms to the feature of the human visual system and the network bandwidth constraints.
Related research results have been published in IUCC 2021.
R. Zhang, X. Zhang*, P. Guo, Q. Fan, H. Yin, and Z. Ma, "QoE Models for Online Video Streaming," 20th International Conference on Ubiquitous Computing and Communications (IUCC/CIT/DSCI/SmartCNS), London, United Kingdom, pp. 32-39, 2021.
• Task 2 proposed neural end-to-end video processing to enhance video quality under bandwidth-limited networks. Specifically, the fellow proposed a residual neural network-based collaborative video processing (CVP) system to trade the computational capability at client-end for QoE improvement via a learned resolution scaling system. Then the fellow proposed a novel raw image-based video analysis framework to perform high-level semantic understanding and low-level compression using RAW images, further reducing the required bandwidth for RGB-based video streaming.
Related research results have been published in IEEE TMC 2023.
W. Miao, G. Min, X. Zhang*, Z. Zhao, J. Hu. Performance Modelling and Quantitative Analysis of Vehicular Edge Computing With Bursty Task Arrivals. IEEE Transactions on Mobile Computing 22(2): 1129-1142 (2023)
• Task 3 designed adaptive video bitrate by using a reinforcement learning-based method, thus accommodating the unexpected network dynamics and guaranteeing users’ QoE. Specifically, the fellow first introduced a lightweight and scalable framework called DMS, for accurate statistical latency measurement on the Internet. Based on the accurate network awareness, the fellow proposed a lightweight Deep Reinforcement Learning (DRL)-based approach to enhance the quality of the video ingestion process of a traffic monitoring system by adjusting the video bitrate adaptively in a real-time manner.
Related research results have been published in IEEE TII 2022 and ACM TOSN 2022.
X. Zhang, G. Min, Q. Fan, H. Yin, D. O. Wu, Z. Ma. A Light-Weight Statistical Latency Measurement Platform at Scale. IEEE Trans. Ind. Informatics 18(2): 1186-1196, 2022.
X. Zhang, Y. Zhao, G. Min, W. Miao, H. Huang, Z. Ma. Intelligent Video Ingestion for Real-time Traffic Monitoring. ACM Trans. Sens. Networks 18(3): 47:1-47:13, 2022.
• Task 4 mainly adopted the results obtained in Task 1 to Task 3 and developed an AI-empowered metaverse prototype for smart cities and beyond. The prototype can provide immersive and personalized experiences to users, thus can accommodate various prospective applications, ranging from smart cities, online education, healthcare, and manufacturing, to real estate.
Related research results have been published in IEEE Communication Magazine 2023.
X. Zhang, G. Min, T. Li, Z. Ma, X. Cao, S. Wang. AI and Blockchain Empowered Metaverse for Web 3.0: Vision, Architecture, and Future Directions. IEEE Communication Magazine 61(8): 60-66, 2023.
