SUPR
User/AP selection strategy design of communication system by combining GNN and reinforcement learning
Dnr:

NAISS 2023/22-307

Type:

NAISS Small Compute

Principal Investigator:

Han Yu

Affiliation:

Chalmers tekniska högskola

Start Date:

2023-03-13

End Date:

2024-04-01

Primary Classification:

20203: Communication Systems

Webpage:

Allocation

Abstract

Dear manager This is Han who is applying the small project of Alvis. I am a postdoc at Chalmers University of Technology's E2 department. My research focuses on the distributed massive input massive output (MIMO) in communication beyond 5G (6G). In this project, I am primarily concerned with the strategy design of user/access point (AP) selection and scheduling. This is a promising communication topic, and due to the advancement of machine learning technology, I would like to apply advanced reinforcement learning and graph neural network (GNN) to address the communication problem in this project. I believe t his is a commendable attempt to apply machine learning to communication. In addition, not only my own idea, but also the EU project Hexa-X-II, we have recently collaborated on the DMIMO area. Following a discussion with the other members of the task I've joined, I've decided to attempt to employ machine learning in this task. Due to the GNN and reinforcement learning model requirements, after I connected with C3SE's support, he kindly advised me that I could apply the Alvis project, as it provides outstanding GPU, CPU, and some specific package resource support. It is true that the dgl package, which I need for my model, is only available for Alvis. If it is possible, I would appreciate you if I could successfully implement this Alvis small project. Many thanks Han