5G • 5G-Advanced (Rel-18)
Over-The-Air (OTA) Retraining
Over-The-Air Neural Network Model Retraining Signaling
Standardized signaling protocol that triggers retraining of active air interface neural networks when environmental conditions shift.
Technical Explanation
When concept drift occurs, base stations must coordinate with UEs to gather fresh channel snapshots and update model weights. Rel-18 establishes OTA Retraining signaling, orchestrating reference signal transmissions, gradient feedback, and validation test benchmarks across the air interface without interrupting active data plane sessions.
Key Functions
- Orchestrates dataset collection periods over the radio interface
- Sends training activation commands and parameter updates over RRC/MAC CE
- Validates newly trained neural network models against standard accuracy KPIs
- Seamlessly transitions live user traffic to newly deployed model checkpoints
Specifications
3GPP TR 38.843, TS 38.331 (Rel-18)
Interfaces
Uu (RRC / MAC CE)
Want to memorize 5G concepts like this one?
Study it with SuperMemo SM-2 spaced repetition flashcards.