公开(公告)号：US20220124543A1

公开(公告)日：2022-04-21

申请号：US17561563

申请日：2021-12-23

Applicant: Oner ORHAN ,  Vasuki NARASIMHA SWAMY ,  Marcel NASSAR ,  Hosein NIKOPOUR ,  Shilpa TALWAR

Inventor： Oner ORHAN ,  Vasuki NARASIMHA SWAMY ,  Marcel NASSAR ,  Hosein NIKOPOUR ,  Shilpa TALWAR

IPC: H04W28/02 ,  H04W24/02 ,  H04W28/08 ,  G06N3/08

Abstract: The present disclosure provides connection management techniques based on graph neural networks (GNN) and deep reinforcement learning (DRL) to optimize user association and load balancing. A graph structure of a communication network is considered for the GNN architecture and DRL is used to learn parameters of the GNN algorithm/model. Connection management is defined as a combinatorial graph optimization problem, and the DRL mechanism uses the underlying graph to learn weights of the GNN for an optimal user connections or associations. The connection management techniques can consider local network features to make better decisions to balance network traffic load while network throughput is also maximized. Implementations are provided based on edge computing frameworks include the Open RAN (O-RAN) architecture. Other embodiments may be described and/or claimed.

公开(公告)号：US20230093673A1

公开(公告)日：2023-03-23

申请号：US17483208

申请日：2021-09-23

Applicant: Vasuki Narasimha Swamy ,  Hosein Nikopour ,  Oner Orhan ,  Shilpa Talwar

Inventor： Vasuki Narasimha Swamy ,  Hosein Nikopour ,  Oner Orhan ,  Shilpa Talwar

IPC: H04W24/02 ,  G06N3/08 ,  G06N3/04

Abstract: A computing node to implement an RL management entity in an NG wireless network includes a NIC and processing circuitry coupled to the NIC. The processing circuitry is configured to generate a plurality of network measurements for a corresponding plurality of network functions. The functions are configured as a plurality of ML models forming a multi-level hierarchy. Control signaling from an ML model of the plurality is decoded, the ML model being at a predetermined level (e.g., a lowest level) in the hierarchy. The control signaling is responsive to a corresponding network measurement and at least second control signaling from a second ML model at a level that is higher than the predetermined level. A plurality of reward functions is generated for training the ML models, based on the control signaling from the MLO model at the predetermined level in the multi-level hierarchy.