SALAD: Self-Adaptive Link Adaptation

TL;DR

SALAD addresses the challenge of achieving high spectral efficiency in wireless downlinks while maintaining a target BLER by inferring SINR solely from ACK/NACK feedback. It combines a teacher–student SINR inference framework with cross-entropy loss, online learning-rate adaptation via knowledge distillation, and a hypothesis-testing-driven MCS probing strategy to rapidly track improving channels. A PID-like feedback loop enforces the long-term BLER target, and SALAD demonstrates robustness to HARQ delays. Through full-stack simulations and OTA 5G experiments, SALAD outperforms the industry-standard OLLA by up to 15\% in throughput and SE across various traffic regimes, while maintaining the BLER target.

Abstract

Adapting the modulation and coding scheme (MCS) to the wireless link quality is critical for maximizing spectral efficiency while ensuring reliability. We propose SALAD (self-adaptive link adaptation), an algorithm that exclusively leverages ACK/NACK feedback to reliably track the evolution of the signal-to-interference-plus-noise ratio (SINR), achieving high spectral efficiency while keeping the long-term block error rate (BLER) near a desired target. SALAD infers the SINR by minimizing the cross-entropy loss between received ACK/NACKs and predicted BLER values, with a learning rate that self-adapts online through knowledge distillation. Based on this inference, SALAD selects the MCS via hypothesis testing: if the SINR is likely underestimated, a higher MCS is selected to accelerate link adaptation under improving channel conditions. To prevent BLER drift from its long-term target, SALAD incorporates a feedback control loop that adjusts the instantaneous BLER target. Over-the-air experiments on a 5G testbed demonstrate that SALAD consistently outperforms the industry-standard outer-loop link adaptation (OLLA). With a single set of parameters, SALAD achieves up to 15% higher throughput and spectral efficiency than multiple OLLA variants across different traffic regimes, while meeting the BLER target.

Figures (8)

• Figure 1: OLLA's behavior when the SINR switches between two values and HARQ feedback delay $\delta=5$. For small values of $\Delta^{\textsc{NACK}}$, the MCS adapts slowly to the channel conditions, and the BLER (over a sliding window of 50 slots) exhibits high variance. For large $\Delta^{\textsc{NACK}}$, OLLA tracks the SINR variations more rapidly but introduces excessive MCS fluctuations in the stationary phase, causing SE degradation. The oracle knows the ground truth. The code used to produce this result is available at ourcode.
• Figure 2: SALAD's main features. High values of the bias score ratio indicate SINR underestimation, triggering the probing of high instantaneous BLER target $\tau_t$ and corresponding MCS indices. This results in a quick rate and SINR adaptation. The instantaneous BLER target $\tau_t$ is adjusted according to the integral error $E_t$, steering the long-term BLER towards its target $\tau$.
• Figure 3: SALAD's workflow diagram.
• Figure 4: Controlled SINR. SINR estimation and MCS index for SALAD and OLLA from OAI RFSim with 7DS2U TDD slot pattern.
• Figure 5: Measurement setup in anechoic chamber at ETH Zürich with COTS UE and rotational absorber table (left) and COTS O-RU (right).