Wireless World Models and Foundation Models

This research area explores how world models and foundation models can provide a shared intelligence layer for wireless networks. Instead of training isolated models for every task, the goal is to learn latent representations that capture how traffic, mobility, radio conditions, energy systems, and sensing data evolve across time.

World models emphasize action-conditioned prediction and planning, while foundation models emphasize reusable representations that transfer across downstream tasks. Together they create a path toward systems that can simulate future states, support digital twins, improve sample efficiency, and enable more autonomous prediction, planning, and control in complex wireless environments.