Massive MIMO Performance Prediction Based on Network Data Learning
- Resource Type
- Conference
- Authors
- Yang, Zhixiong; Krishnakumar, Poornima; Hosseini, S. Amir; Ng, Chris
- Source
- 2021 55th Asilomar Conference on Signals, Systems, and Computers Signals, Systems, and Computers, 2021 55th Asilomar Conference on. :1255-1262 Oct, 2021
- Subject
- Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Computing and Processing
Fields, Waves and Electromagnetics
Signal Processing and Analysis
Computers
Machine learning algorithms
Spectral efficiency
Decision making
Optimization methods
Massive MIMO
Machine learning
- Language
- ISSN
- 2576-2303
Beamforming is one of the key challenges to improve the performance of Massive MIMO networks. This work focuses on beam optimization where the base station (BS) will select the best beam from a beam library to improve the spectral efficiency and maintain high volume. Machine learning approaches are adopted to accomplish two tasks: Reference Signal Received Power (RSRP) prediction and volume prediction. By observing the network data from a small number of beams, the BS can learn to predict RSRP and traffic distributions in the given environment. Then an optimization algorithm will select the best beam combinations based on the predictions.