Deep Reinforcement Learning for Efficient IoT Data Compression in Smart Railroad Management
Xuan Chen, Qixuan Yu, Shuhong Dai, Pengfei Sun, Haichuan Tang, Long Cheng
Abstract
Modern smart railroad management relies heavily on Internet of Things (IoT)-enabled sensors to monitor train performance, resulting in the generation of extensive data streams. The ensuing data influx poses critical challenges in storage, processing, and transmission. This paper presents a novel data compression method specifically designed for IoT-generated data within railroad management scenarios. Leveraging deep reinforcement learning (DRL), our approach intelligently compresses data from onboard IoT sensors. This not only streamlines data streams, ensuring essential information is retained and redundant data is pruned, but also alleviates strain on resource-limited devices such as edge servers tasked with complex computations or aggregating large datasets for long-term trend analysis. Experimental results demonstrate that our approach can outperform current baselines with achieving an enhancement of more than 18% on compression rate in real-time configurations. This signifies a transformative solution for handling IoT-induced big data in contemporary railroad management systems.