AI beam selection for 5G mmWave

The brief

5G millimetre-wave links need the base station to point a narrow beam at each user, and sweeping every beam wastes time. Using ray-traced channel data, train a model that predicts the best beams from a user's position or a few measurements. Good looks like: top-k beam accuracy close to an exhaustive sweep at a fraction of the overhead, with an analysis of when the model fails.

What you will build

1. Generate or load channel data for a scenario
2. Set up the beam codebook and the exhaustive-sweep baseline
3. Train models to predict the best beams
4. Measure top-k accuracy and achievable rate
5. Study sensitivity to position errors and blockage
6. Write the results up like a research paper

Free, open data to build it with

• DeepMIMO ray-traced channel datasets ↗ DeepMIMO (Arizona State University)

Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.