5G • 5G-Advanced (Rel-18)
AI/ML MU-MIMO Pairing
Deep Graph Neural Network Multi-User MIMO User Scheduling
Employs Graph Neural Networks (GNN) to solve combinatorial user pairing and precoder selection for Massive MIMO in sub-millisecond scheduling intervals.
Technical Explanation
Pairing co-scheduled UEs in Massive MIMO requires evaluating orthogonality between dozens of candidate spatial channels—an NP-hard combinatorial problem that overwhelms gNodeB DSPs. In Rel-18, Graph Neural Networks (GNNs) model spatial correlations between active users, finding near-optimal user subsets and zero-forcing precoders in real time.
Key Functions
- Solves high-complexity combinatorial user selection in under 0.5 ms
- Maximizes sum-rate capacity while maintaining fairness across cell-edge users
- Minimizes inter-user spatial interference through optimal null-space steering
- Scales efficiently to 64-TRX and 128-TRX Massive MIMO base station configurations
Specifications
3GPP TR 38.843, TS 38.214 (Rel-18)
Interfaces
UuMAC Scheduler
Related 5G Concepts
Want to memorize 5G concepts like this one?
Study it with SuperMemo SM-2 spaced repetition flashcards.