Autoencoder for Position-Assisted Beam Prediction in mmWave ISAC Systems

2025-11-23

AI Summary

In millimeter-wave integrated sensing and communication (ISAC) systems, narrow beams incur substantial training overhead, while conventional fully connected neural networks suffer from high computational complexity. To address these challenges, this paper proposes a lightweight autoencoder-based beam prediction method that explicitly incorporates user location priors. We design a three-layer undercomplete autoencoder architecture that jointly encodes spatial position information during feature compression and reconstruction, enabling end-to-end, low-complexity beam alignment. Compared to a baseline fully connected network, the proposed model reduces computational cost by 83% while maintaining comparable beam prediction accuracy. This approach achieves an effective trade-off between precision and efficiency, offering a deployable, lightweight solution for real-time beam management in 6G millimeter-wave ISAC systems.

Technology Category

• Machine Learning: Hardware-aware ML
• Search and Optimization: Learning to Search
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

Integrated sensing and communication and millimeter wave (mmWave) have emerged as pivotal technologies for 6G networks. However, the narrow nature of mmWave beams requires precise alignments that typically necessitate large training overhead. This overhead can be reduced by incorporating the position information with beam adjustments. This letter proposes a lightweight autorencoder (LAE) model that addresses the position-assisted beam prediction problem while significantly reducing computational complexity compared to the conventional baseline method, i.e., deep fully connected neural network. The proposed LAE is designed as a three-layer undercomplete network to exploit its dimensionality reduction capabilities and thereby mitigate the computational requirements of the trained model. Simulation results show that the proposed model achieves a similar beam prediction accuracy to the baseline with an 83% complexity reduction.

Problem

• Predicting mmWave beam alignment using position information
• Reducing computational complexity in beam prediction models
• Achieving high accuracy with lightweight autoencoder design

Innovation

• Lightweight autoencoder for position-assisted beam prediction
• Three-layer undercomplete network reduces computational complexity
• Achieves similar accuracy with 83% complexity reduction