Logo image
Unified Packet Compression and Model Adaptation for Integrated Sensing and Multi-Modal Communications
Journal article   Peer reviewed

Unified Packet Compression and Model Adaptation for Integrated Sensing and Multi-Modal Communications

Xuanhao Luo, Zhouyu Li, Mingzhe Chen, Ruozhou Yu, Shiwen Mao and Yuchen Liu
IEEE journal on selected areas in communications, Vol.44, pp.1-1
2025-09-16

Abstract

Adaptation models Byte-based model Computational modeling Data communication Data compression Data models Entropy Image coding multi-modal sensing and communications packet compression Predictive models Redundancy Transformer Transformers
Integrated sensing and communication systems face critical challenges, including limited bandwidth, power constraints, and varying communication conditions, which demand efficient data transmission and processing strategies. This paper introduces, ByteTrans, a novel joint optimization framework that integrates byte-level predictive modeling with adaptive model scheduling to maximize data transmission efficiency while adhering to communication and computational constraints. The proposed framework employs Transformer-based models to predict and compress data packets losslessly, leveraging the inherent redundancy in multi-modal network data. Such a unified data compression approach predicts occurring byte probabilities, encodes them as ranks using lossless entropy coding, and efficiently reduces data size and entropy across diverse modalities. Then, a dynamic adaptation strategy selects the optimal compression model based on packet characteristics and channel conditions, ensuring efficient operation across heterogeneous sensor environments. Experimental results validate that our scheme achieves compression rates exceeding 50%, while showcasing substantial reductions in communication time and bandwidth usage under both normal and adverse channel conditions. Furthermore, we effectively implement these models across various real-world edge sensors and servers, showcasing their practicality and efficiency in various network applications. By addressing the trade-offs between achieving lower compression ratios and limiting computational and energy consumption, this work establishes a scalable and robust solution for data management in multi-modal communication systems.

Metrics

18 Record Views

Details

Logo image