It is envisioned that 6G will extend beyond mobile Internet, supporting ubiquitous intelligence services and Internet of Everything (IoE) applications. These applications not only include the requirements to guarantee classic end-to-end communications (e.g., massive MIMO channel estimation and beamforming) but also cover broader scenes such as sustainable cities, connected autonomous systems, digital twins, and metaverse. Their unique requirements and goals pose critical challenges to traditional Shannon’s bit-oriented communication frameworks, which are nearing the Shannon capacity limit. To enhance communication efficiency, recent studies in semantic communications have focused on transmitting information with semantic meaning while excluding insignificant data. However, these transmitter-centered techniques cannot effectively characterize the significance of data in specific tasks, as the significance of the same data varies across different tasks and over time on the receiver side. To address this issue, task-oriented and generative communications have recently attracted considerable attention. Task-oriented communications consider the effective and efficient completion of specific tasks at the receiver side, whereas generative communications activate communications only when receivers cannot address task and data significance variations by locally generating desired data. In particular, utilizing recently developed deep generative models (DGMs) such as ChatGPT and stable diffusion models, receivers can generate high-fidelity target content by receiving very little key information for a specific task (e.g., using very little text information to generate a high-resolution image), while being able to flexibly change data semantics (e.g., pose and partial objects in images) and modalities (e.g., from text to a video clip). Thus, task-oriented and generative communications have a synergy to meet the requirements of intelligence services, especially for time-varying and generic tasks without compromising communication efficiency.

However, there are still many fundamental problems in task-oriented and generative communications that remain under-explored. Observing the recent surge in related research and the emergence of exciting opportunities, we propose the timely and promising workshop “Task-Oriented and Generative Communications for 6G.” This workshop aims to bring together researchers from both academia and industry to present the latest advancements and discuss future research opportunities in this field. 

Topics of interest include, but are not limited to: 

•  Information theory for task-oriented and generative communications

•  Signal processing for task-oriented and generative communications

•  New architecture for task-oriented and generative communications

•  New protocols for task-oriented and generative communications

•  Task-oriented joint source and channel coding

•  Large language model (LLM) based deep joint source and channel coding

•  Multimodal generative model based deep joint source and channel coding

•  Generative model based channel emulation and estimation

•  Task-oriented and generative model based quantization

•  Task-oriented distributed learning over wireless systems

•  Distributed generative model fine-tuning over wireless systems

•  Task-oriented and generative techniques for edge AI

•  Task-oriented and generative techniques for integrated sensing, communication, and computation

Paper Submission Link

https://edas.info/newPaper.php?c=31746&track=121199

Important Dates

Workshop Paper Submission: 04 February 2024

Paper Acceptance Notification: 06 March 2024

Camera Ready: 15 March 2024 

Registration Due for Accepted Papers: 15 March 2024