MAC Scheduler

The scheduler's primary function is to map user data to the time-frequency resource grid every Transmission Time Interval (TTI). It assigns Physical Resource Blocks (PRBs) to User Equipment (UE) based on instantaneous demand and channel conditions. This is not a static assignment; it is a dynamic, per-slot decision that adapts to the fading dips and peaks of each user's radio link, maximizing the utilization of the available spectrum.

Modern schedulers exploit multi-user diversity by prioritizing users when their channel quality is high. The scheduler ingests Channel Quality Indicator (CQI) reports from UEs to estimate the supportable data rate for each PRB. Key strategies include:

• Max C/I: Allocates resources to the user with the best instantaneous channel, maximizing cell throughput at the cost of fairness.
• Proportional Fair (PF): Balances throughput and fairness by scheduling users based on the ratio of their instantaneous rate to their past average throughput.

The scheduler is the enforcement point for Quality of Service (QoS). It manages distinct bearer types with different priorities:

• Guaranteed Bit Rate (GBR): For services like VoIP, the scheduler must ensure a minimum bit rate is met before allocating resources to non-GBR flows.
• Non-GBR: For best-effort web traffic, resources are shared dynamically.
• Delay-Critical GBR: For ultra-reliable low-latency communication (URLLC), the scheduler must prioritize transmissions to meet a strict packet delay budget, often using pre-emption over other traffic.

The scheduling decision is tightly coupled with the Multiple-Input Multiple-Output (MIMO) configuration. The scheduler must decide not only which PRBs to assign but also how many spatial layers (or streams) to use. It selects between transmit diversity for robust, low-SINR connections and spatial multiplexing for high-SINR, high-throughput scenarios. This involves pairing users on the same time-frequency resource for Multi-User MIMO (MU-MIMO) to increase cell capacity without additional spectrum.

The scheduler works in a closed loop with Link Adaptation and Hybrid Automatic Repeat Request (HARQ). It selects a Modulation and Coding Scheme (MCS) based on the predicted channel quality. An aggressive MCS yields high throughput but risks a failed transmission. The scheduler must account for pending HARQ retransmissions, which often take strict priority over new data to clear the HARQ buffer and avoid upper-layer timeouts, directly influencing the effective user throughput.

A MAC scheduler operates under an extreme real-time constraint. The entire decision cycle—from collecting buffer status reports and CQIs to computing the allocation matrix—must complete within a fraction of the TTI (e.g., < 100 µs for a 1 ms TTI in 5G). This requires highly optimized algorithms implemented in dedicated hardware accelerators or tightly coupled software on bare-metal CPUs. The complexity of the algorithm is directly bounded by this processing deadline.

The primary input to any frequency-selective scheduler. CSI reports—including CQI, PMI, and RI—tell the scheduler which resource blocks are viable for each user. In 5G NR, the scheduler relies on SRS-based channel reciprocity in TDD mode. AI/ML-based CSI prediction aims to forecast channel aging, allowing the scheduler to make decisions based on future channel conditions rather than stale measurements, dramatically improving spectral efficiency at high mobility.

The scheduler's constraint system. Each data radio bearer maps to a 5QI (5G QoS Identifier) with specific requirements:

• GBR flows: Guaranteed bit rate, e.g., conversational voice (5QI=1)
• Non-GBR flows: Best-effort, e.g., web browsing (5QI=6/8/9)
• Delay-critical GBR: Ultra-reliable low-latency (5QI=82-85) The scheduler must satisfy GBR guarantees first, then distribute remaining resources proportionally among non-GBR flows using weighted fair queuing or similar algorithms.

Defines how UE positions evolve during simulation, directly impacting the scheduler's handover and resource allocation logic. Common models include:

• Random Waypoint: UEs pick random destinations and speeds
• Gauss-Markov: Smooth, temporally correlated movement
• Trace-based: Replayed GPS logs from real drive tests High mobility (e.g., vehicular at 120 km/h) stresses the scheduler with rapid channel quality fluctuations and frequent handover triggers, testing the robustness of CSI aging compensation.

The canonical algorithm balancing throughput maximization and user fairness. The metric is: M_i = R_i(t) / T_i(t), where R_i(t) is the instantaneous achievable rate (from CSI) and T_i(t) is the exponentially averaged past throughput. Users with a high instantaneous rate relative to their history get priority. This exploits multi-user diversity—scheduling users when their channel peaks—while preventing starvation of cell-edge users. Variants include adaptive PF that adjusts the averaging window based on delay sensitivity.

The scheduler's companion function that selects the Modulation and Coding Scheme (MCS) for each allocation. Based on the reported CQI, link adaptation chooses an MCS that maximizes throughput while maintaining a target Block Error Rate (BLER), typically 10% for initial transmissions. Outer loop link adaptation dynamically adjusts the MCS offset based on HARQ ACK/NACK statistics, compensating for CQI measurement inaccuracies. The scheduler and link adaptation loop jointly determine the final transport block size.