3rd Workshop on
Semantic Communication for 6G
Organised in Conjunction with IEEE GLOBECOM 2023
Recently there has been huge interest in semantic communication (SC) alongside a lively academic debate on defining what SC will be, as evidenced by the visual summary below.
Notwithstanding, many of the current standardization activities and industry initiatives have not yet been considering SC as their 6G candidate technologies. To fill this gap and usher in 6G SC systems, this forum aims to facilitate in-depth discussions on the current/future research directions as well as on how to unify/distinguish different points of view, by answering the following questions.
AI/ML-native and human/machine-centric SC
Q1. More than several SC visions are aligned with another emerging research direction of AI/ML-native communication. Compared to AI/ML-native communication, what are new fundamental challenges and possible research topics in SC?
Q2. The current SC frameworks often rest on human-centric visual/auditory perception, language, and/or cognition. What can be new principles and approaches for machine-centric SC?
Q3. Recent AI/ML models can generate high-quality image/audio/video samples based simply on human language instructions, enabling cross-modal compression/decompression to/from human languages. Will SC be the technical communication of human languages or of machine/task-specific new languages?
*The cover image of this webpage was generated by OpenAI’s DALLE-E with the human-language prompt: “a centered explosion of colorful particles on a black background, in the shape of 6G.”
6G-ready SC systems and technologies
Q4. The current SC technologies are far from large-scale networks and system-level operations. How can we integrate SC into which layers of the existing network architectures?
Q5. Unlike the current general-purpose communication systems, SC is task-oriented and agent-specific, which may not guarantee universal connectivity. How can we develop compatible and scalable SC systems?
Q6. ITU WRC 2023 and LA 2028 Summer Olympics are expected to be two major events shaping what 6G will be. Considering these short and long-term milestones, which SC techniques can be put forward as 6G candidate technologies?
Call for Papers
Scope and Topics
The traditional approach of “the higher the better, and the more the merrier” in communication engineering is no longer sustainable for 6G due to hardware limitations and increasing networking costs. Alternatively, recent advances in artificial intelligence (AI) and on-device computing technologies offer new opportunities for intelligent devices at the edge of 6G networks. Semantic communication (SC) is an emerging paradigm of communication and networking that leverages these capabilities, allowing distributed AI-native edge devices to maximize their task-specific effectiveness by understanding and manipulating the semantics of transferred bits. While promising, SC is in its infancy, and faces significant challenges. Existing SC studies often focus on PHY layers in point-to-point scenarios, ignoring 6G architecture compatibility. MAC and higher layers for SC as well as large-scale network issues remain still unexplored. Furthermore, arguably due to the ill-defined notion of semantics, the current SC frameworks are severely fragmented. Towards shaping what 6G SC will be, this workshop solicits novel works on SC and their networking and edge computing issues.
Topics of interest include, but are not limited to:
Deep learning based SC frameworks and applications (e.g., DeepSC and DeepJSCC)
Timing based SC analysis and applications (e.g., age of incorrect information, urgency of information, query age of information, and value of information)
Knowledge base assisted SC frameworks and applications (e.g., knowledge graphs, structured causal models, probabilistic logic, and GFlowNet)
Game-theoretic and reinforcement learning based SC frameworks and applications (e.g., signaling games, rational speech act models, coordinative multi-agent reinforcement learning)
Communication/machine learning principles and mathematical theories for SC (e.g., source-oriented rate-distortion theory, variational information bottleneck principles, and enriched category theory)
Novel metrics for evaluating semantic information and SC performance (e.g., semantic entropy, causal influence of communication, context independence, and structural intervention distance)
Semantic MAC protocols and applications (e.g., protocol learning, causal and symbolic protocols)
SC network architectures and cross-layer designs (e.g., IE-SC, FEI, and HCP)
AI-native/model-based networking frameworks for SC (e.g., protocol learning and intent-based networking)
Human/machine-centric and human-machine interactive SC frameworks and applications
Experimental/simulation designs for SC and use cases
Important Dates
• Submission deadline: July 15, 2023
• Notification of acceptance: September 15, 2023
• Camera Ready: October 7, 2023
Organizers
General Chairs:
Jihong Park, Deakin University, Australia
Jinho Choi, Deakin University, Australia
Seong-Lyun Kim, Yonsei University, Korea
Seung-Woo Ko, Inha University, Korea
Zhijin Qin, Tsinghua University, China
Contact: Jihong Park (jihong.park at deakin.edu.au)
Link to the 1st workshop on SC6G at APCC 2022: https://sites.google.com/view/sc6g
Link to the 2nd workshop on SC6G at IEEE SECON 2023: https://sites.google.com/view/sc6g-secon23
Link to the BrainLink workshop on SC6G 2023: https://sites.google.com/view/sc6g-brainlink
Program
Session 1 - Chair: Jihong Park (9:00-10:30)
A Wyner-Ziv Coding-Based Semantic Communication Approach with a Shared Semantic Codebook
Deep Learning Enabled Video Semantic Transmission Against Multi-dimensional Noise
Multiuser Resource Allocation for Semantic-Relay-Aided Text Transmissions
Optical Flow-Based Video Sketch Graph Extraction
Neural Estimation for Rate-Distortion Function in Semantic Communications
Energy-Efficient Downlink Semantic Generative Communication with Text-to-Image Generators
Break and Discussion (10:30-11:00)
Session 2 - Chair: Jinho Choi (11:00-12:30)
MDVSC---Wireless Model Division Video Semantic Communication for 6G
Semantic Communications for Joint Image Recovery and Classification
Resource Allocation for Semantic-Aware Mobile Edge Computing Systems
Semantic Multi-Resolution Communications
Optimal Semantic-aware Sampling and Transmission in Energy Harvesting Systems Through the AoII
Spatiotemporal Attention-based Semantic Compression for Real-time Video Recognition
Concluding Remarks - Seong-Lyun Kim