Background

Modern wireless networks extensively rely on multi-user, multiple-input, multiple-output (MU-MIMO) technology. In MU-MIMO technology, to correctly beamform transmissions, the beamformer needs to frequently acquire a beamforming matrix (BM) from each connected beamformee. However, the size of the BM grows with the number of antennas and subcarriers resulting in an increased amount of airtime and computational load at the beamformee. Conventional approaches come with either excessive computational load or loss of beamforming precision.

Technology Overview

Northeastern researchers have created SplitBeam, a new data-driven framework for Wi-Fi networks which involves training a split deep neural network (DNN) to directly output the BM given the channel state information (CSI) matrix as input. The DNN is designed with an additional “bottleneck” layer to “split” the original DNN into a head model and a tail model, respectively executed by the station and the access point. The head model generates a compressed representation of the BM, which is then used by the AP to produce the BM using the tail model. The bottleneck introduces a trade-off between computation/feedback overhead and reconstruction accuracy. Thus, they can formulate and solve a bottleneck optimization problem to keep computation, airtime overhead, and bit error rate below application requirements. The researchers have performed extensive experimental CSI collection with off-the-shelf Wi-Fi devices in two distinct environments and compared the performance of SplitBeam with the standard IEEE 802.11 algorithm for BM feedback and the state-of-the-art DNN-based approach LB-SciFi. These experimental results show that SplitBeam reduces the beamforming feedback size and computational complexity by respectively up to 81% and 84% while maintaining a bit error rate within about 10-3 of existing approaches.

Benefits

Reduction in beamforming feedback size by up to 81%
Reduction in computational complexity by up to 84%
Enables using higher bandwidth and serving higher number of users
Outperforms existing Wi-Fi protocol and other DNN-based approaches

Applications

Enables next-generation Wi-Fi networks like IEEE 802.11 (Wi-Fi 7 routers)
Opens a new line of research as QoS-aware DNN models
High Client-Dense Wi-Fi environments
Can be used in the deployment of interactive applications

Opportunity

Seeking licensee and/or industry partner

SplitBeam: Rethinking Beamforming Feedback in MU-MIMO Wi-Fi Systems Through Split Neural Networks

IP Status

Seeking