Adaptive Modulation and Coding (AMC)

AMC systems continuously vary the modulation scheme based on the Signal-to-Noise Ratio (SNR). When channel conditions are favorable (high SNR), the transmitter switches to higher-order modulation like 64-QAM or 256-QAM to pack more bits per symbol.

• QPSK: Robust, used in poor channel conditions (low SNR)
• 16-QAM: Moderate spectral efficiency for average conditions
• 64-QAM: High throughput for strong signal environments
• 256-QAM: Maximum spectral efficiency, requires excellent SNR

The transition thresholds are determined by the target Block Error Rate (BLER) , typically 10% for initial transmissions.

Alongside modulation, AMC adjusts the forward error correction (FEC) coding rate to add redundancy proportional to channel impairment. A lower code rate adds more parity bits for error resilience at the cost of reduced net data rate.

• Rate 1/2: 50% redundancy, maximum error protection
• Rate 2/3: Moderate protection with improved throughput
• Rate 3/4: Light protection for good channel conditions
• Rate 5/6: Minimal overhead, near-maximum throughput

Modern systems like 5G NR use LDPC codes for data channels and polar codes for control channels, enabling fine-grained rate matching through puncturing and repetition.

AMC relies on a closed-loop feedback mechanism where the receiver estimates channel quality and reports it back to the transmitter. The Channel Quality Indicator (CQI) is a quantized metric that maps directly to a recommended modulation and coding scheme (MCS) index.

• Receiver: Estimates SNR, SINR, or mutual information per sub-band
• CQI Reporting: Periodic or aperiodic feedback via uplink control channels
• MCS Selection: Transmitter maps CQI to a pre-defined MCS table
• Latency Constraint: Feedback must arrive within the channel coherence time to remain valid

In massive MIMO systems, CSI acquisition becomes more complex, requiring compressed feedback or reciprocity-based estimation in TDD mode.

Wireless standards define discrete Modulation and Coding Scheme (MCS) tables that map integer indices to specific modulation-order and code-rate combinations. 5G NR supports multiple MCS tables optimized for different use cases.

• Table 1: Up to 64-QAM, baseline spectral efficiency
• Table 2: Up to 256-QAM, high throughput for eMBB
• Table 3: Up to 64-QAM, low spectral efficiency for URLLC reliability
• Table 4: Up to 256-QAM, low code rates for high reliability

Each MCS index specifies a spectral efficiency value in bits/s/Hz, allowing the scheduler to select the highest throughput option that satisfies the target BLER constraint.

Pure CQI-based AMC suffers from estimation errors and reporting delays. Outer Loop Link Adaptation applies a dynamic offset to the reported CQI based on observed Hybrid Automatic Repeat Request (HARQ) statistics.

• ACK/NACK Tracking: Monitors first-transmission success/failure rates
• Offset Adjustment: Increments offset on NACK, decrements on ACK
• Convergence Target: Drives actual BLER toward the configured target BLER
• Step Size Tuning: Balances convergence speed against stability

OLLA compensates for systematic biases in CQI estimation, residual interference, and receiver implementation losses, ensuring robust link performance in non-ideal field conditions.

Traditional threshold-based AMC struggles in complex interference environments. Deep reinforcement learning and neural network classifiers can learn optimal MCS selection policies directly from raw channel observations without explicit SNR thresholds.

• DQN-based Selection: Learns state-action mappings from experienced throughput and BLER rewards
• LSTM Predictors: Forecast channel evolution to preemptively select MCS before degradation
• Contextual Bandits: Incorporate additional features like Doppler spread and delay spread
• Transfer Learning: Reuse policies trained in simulation for rapid deployment in real hardware

These AI-driven approaches outperform static lookup tables in high-mobility scenarios and non-stationary interference environments typical of cognitive radio and defense applications.

The Modulation and Coding Scheme index is the discrete output of the AMC algorithm, defining a specific combination of modulation order and code rate. In 5G NR, MCS tables map directly to spectral efficiency. For example:

• MCS 0: QPSK with low code rate for cell-edge users
• MCS 15: 64QAM with high code rate for good conditions
• MCS 27: 256QAM for excellent SINR The AMC engine selects the MCS that maximizes throughput while maintaining a Block Error Rate (BLER) below a target threshold, typically 10%.

Hybrid ARQ is the complementary error-control mechanism that works in tandem with AMC. While AMC provides coarse rate adaptation based on long-term channel statistics, HARQ handles fast-fading variations through incremental redundancy. If a packet fails despite AMC's selection, the receiver stores the soft bits and requests a retransmission. The Chase combining or incremental redundancy gain allows the system to operate at a higher target BLER, enabling AMC to select more aggressive MCS levels for greater spectral efficiency.

SINR is the primary physical-layer metric that AMC algorithms map to specific MCS entries. It quantifies the ratio of desired signal power to the sum of interference and thermal noise. AMC implementations maintain a link curve—a lookup table mapping SINR thresholds to MCS indices. For instance, QPSK may require 0 dB SINR, while 64QAM demands 15 dB. In OFDM systems, SINR is measured per subcarrier, and the AMC engine often uses the effective exponential SNR mapping (EESM) to compress wideband SINR into a single metric.

Outer Loop Link Adaptation is an enhancement to standard AMC that corrects for systematic estimation errors. The receiver tracks the actual Block Error Rate (BLER) and applies a dynamic offset to the SINR estimate. If the BLER exceeds the target, OLLA adds a negative backoff, forcing AMC to select a more robust MCS. If the BLER is too low, a positive offset encourages more aggressive modulation. This closed-loop correction compensates for CSI aging, imperfect calibration, and mobility-induced Doppler shifts.

The water-filling algorithm is the information-theoretic foundation for adaptive modulation in OFDM systems. It allocates more power and higher-order modulation to subcarriers with favorable channel gains while reducing or nulling transmission on deeply faded subcarriers. In practice, AMC approximates water-filling by assigning different MCS levels to different resource blocks based on their individual SINR. This frequency-selective scheduling maximizes the total channel capacity under a total power constraint.