Knowledge Distillation for Lightweight Multimodal Sensing-Aided mmWave Beam Tracking

TL;DR

This paper introduces a knowledge-distillation framework to develop lightweight models for mmWave beam prediction using multimodal sensor data, achieving high accuracy with reduced complexity.

Contribution

It proposes a novel KD-based approach to transfer knowledge from a complex CNN-GRU teacher to a compact student model for multimodal beam tracking.

Findings

Student model achieves over 96% Top-5 accuracy.

Reduces computational complexity by over 4 times.

Decreases number of parameters by over 27 times.

Abstract

Beam training and prediction in real-world millimeter-wave (mmWave) communications systems are challenging due to rapidly time-varying channels and strong interference from surrounding objects. In this context, widely available sensors, such as cameras and radars, can capture rich environmental information, enabling efficient beam management. This paper proposes a knowledge-distillation (KD)-enabled learning framework for developing lightweight and low-complexity models for beam prediction and tracking using real-world camera and radar data from the DeepSense 6G dataset. Specifically, a powerful teacher network based on convolutional neural networks (CNNs) and gated recurrent units (GRUs) is first designed to predict current and future beams from historical sensor observations. Then, a compact student model is constructed and trained via KD to transfer the predictive capability of the teacher model to a lightweight architecture. Simulation results demonstrate that jointly leveraging radar and image modalities significantly outperforms single-modality approaches. Moreover, the proposed student model achieves over 96% Top-5 beam prediction accuracy while reducing computational complexity by more than 4 times and the number of parameters by over 27 times compared with the teacher model.