New & Events

SUNRISE-6G Latest Contributions at the 3GPP SA WG2 Meetings

Lenovo has been at the forefront of innovation, advancing AI and Network Solutions, presenting significant contributions at recent 3GPP SA WG2 meetings, aiming to push the boundaries of AI and network technology.

Here’s a comprehensive overview of our latest innovative proposals:

Vertical Federated Learning Solution

At SA WG2 Meeting #162 in Changsha, China, Lenovo introduced the “Solution for KI#2: Support for Vertical Federated Learning (VFL): Model Training and Inference” as part of the FS_AIML_CN work item for Rel-19.

Key Features:

Sample and Feature Alignment: Ensures all VFL participants are synchronised before the federated learning process begins.
Coordinated Training: Manages active and passive participants for optimal data use and model accuracy.
Distributed Inference: Facilitates efficient, scalable inference leveraging multiple VFL participants.

This contribution aligns with SUNRISE-6 G’s goal of employing state-of-the-art transfer learning, including federated learning and continual learning techniques. It proposes procedures that can enable vertical federated learning across different testbeds.

DualSteer Solution Based on ATSSS

Lenovo also proposed (at SA WG2 Meeting #162) the “DualSteer Solution Based on ATSSS” as part of the FS_MASSS work item for Rel-19.

Highlights:

Dual 3GPP Access Paths: Supports MA PDU Sessions with two 3GPP paths, and optionally, one non-3GPP path.
Minimal Network Impact: Seamless integration without extensive 5G registration changes.
Versatile Connectivity: Utilizes MPTCP or MPQUIC for robust data transmission.

This solution supports SUNRISE-6 G’s aim to enable multi-access connectivity using terrestrial (NR) and non-terrestrial (NR-NTN) access paths. The proposed new multi-access solution supports multi-access connectivity using two 3GPP access paths.