Machine Learning-Enhanced Data Transmission for Autonomous Driving and IoT

Abstract

Managing the exponential growth of data, particularly video and sensor data, is a significant challenge in cloud and IoT environments. This research introduces a novel AI-driven data offloading technique aimed at minimizing latency and optimizing resource utilization in cloud computing systems. By leveraging advanced machine learning models including frame overlap detection, recurrent neural networks, and transformers the proposed approach delivers substantial performance improvements. Experimental results indicate a 28.26% reduction in average latency, a 22.96% decrease in cloud resource utilization, and a 41.66% reduction in bandwidth consumption. Additionally, the novel radial compression method achieved a final compression rate of 33.33%. The technique dynamically adjusts compression levels based on network conditions, intelligently identifying and transmitting only essential data elements. Validation using the KITTI autonomous driving dataset demonstrates its potential to enhance data transmission efficiency in real-world IoT applications. This study highlights the effectiveness of AI-powered strategies in addressing the increasing demands of data offloading in modern cloud-IoT ecosystems.