Traffic prediction for Backhaul Networks Using AI

Purpose and goal

Wireless backhaul networks offer many advantages over optical fibre, but one drawback is the comparatively large energy consumption. Accurate prediction of future capacity demands is an enabler for energy savings. This can be achieved in many ways, for example, by adjusting the output power to cater to instantaneous traffic. The purpose of this project is to use AI tools to predict long term traffic patterns over wireless backhaul links.

Expected effects and result

The expected output of this project is an AI based traffic prediction model adapted to various classess of wireless backhaul links. The model will be trained using existing traffic data to achieve substantial prediction accuracy. The AI based model may be deployed in backhaul links wherein the link can use the predicted traffic demand to optimize power usage.

Planned approach and implementation

The project plan involves various steps, starting with gathering the required data from existing data lakes within the company. Then we use off the shelf classification tools to classify the links based on traffic arrival patterns. Then for each of these classes we develop AI based prediction models. The prediction models will be optimized to have very high accuracy. The project will be implemented in a python based environment.