MUSE-COM ² objective is to develop the AI-empowered multimodal communication considering semantics of individual modalities jointly managed and optimized with information processing in edge servers.
Following specific objectives are defined:
Architecture
Define architecture for joint semantic communications and computing framework, including integration of AI and its interaction with traditional network elements, and adjust it for targeted industrial use-cases
Analysis of modalities
Determine modalities to be transferred and manage communication and computing for individual modalities to minimize the amount of data to be transferred to reduce load of communication links and load of MEC servers.
AI training optimization
Develop a low-energy consumption method for AI training of large datasets required to select modalities to be transferred and processed
AI-driven computing and communication resource allocation
Develop AI-driven joint computing and communication resource management algorithm to deliver individual modalities to MEC server
Implementation and testing in lab
Implement and test the developed solutions for communication and MEC processing in an open source SW-defined radio testbed in the lab
Real-world PoC
Validate the proposed concept and solutions via experiments in the real-world environment in a facility of BOSCH for a realistic industrial use-case
Business aspects
Outline business cases and provide business analysis for the application of the developed solution in industry and define use-cases and applications in other domains and verticals
Dissemination
Disseminate the project results by the project partners to various communities and fora considering open science principles
We focus on the scenario, where a local multimodal information of a device is offloaded and processed remotely in the edge servers. Each modality carries specific information to be treated in different ways due to its impact on and importance for the results of MEC processing.
MUSE-COM² targets to validate the concept via two use-cases:
Industrial use case
Vehicular use case
