Age-Optimal Target Wake Time: Provably Good Wake Schedules for Energy-Constrained Wi-Fi Status Updating

What the math gives to ML

The paper gives an exact renewal-reward model for the age of unreliable periodic updates and a convex water-filling rule for allocating heterogeneous refresh intervals. This structure transfers to federated learning, sensor-conditioned world models, and retrieval systems where data sources differ in importance, reliability, and communication cost. A scheduler can refresh important reliable sources frequently while slowing expensive or unreliable sources, then pack the resulting requests into non-overlapping communication slots. The neural architecture need not change, so the proposal is suitable for a direct systems experiment.