Machine Learning for Throughput Prediction in Multi-AP Wi-Fi Networks using Coordinated Spatial Reuse
* Register (or log in) to the AI4G Neural Network to add this session to your agenda or watch the replay
Multi-Access Point Coordination (MAPC) is one of the new features to be included in next generation Wi-Fi networks. It represents a radical change in the way traditional Wi-Fi networks work. Following the MAPC framework, APs can agree on how to share spectrum resources, aiming to improve the network efficiency by reducing unnecessary contention inside the same WLAN network. Several schemes like coordinated Spatial Reuse (c-SR), where multiple APs can transmit simultaneously, have been proposed to take advantage of the MAPC framework.
This problem statement proposes the use of Machine Learning (ML) to predict the throughput that a subset of APs using c-SR can achieve. Predicting the performance of different sets of transmitting devices is essential to derive the optimal set of groups in WLAN settings. Then, our goal is to build an ML model able to predict the achievable throughput given any random -and maybe not present in the original dataset- combination of APs that aim to transmit simultaneously and also use it to find the best groups of APs in different network scenarios, and topologies.