Empirical evaluation of predictive channel-aware transmission for resource efficient car-to-cloud communication

TL;DR

This paper empirically evaluates three car-to-cloud data transmission algorithms, focusing on resource efficiency by leveraging channel quality and predictive modeling in LTE networks.

Contribution

It introduces and compares empirical performance of channel-aware, predictive, and periodic transmission schemes for vehicle data uploads.

Findings

pCAT outperforms periodic scheme in resource efficiency

Channel quality measurements improve upload timing

Predictive modeling enhances transmission decisions

Abstract

Nowadays vehicles are by default equipped with communication hardware. This enables new possibilities of connected services, like vehicles serving as highly mobile sensor platforms in the Internet of Things (IoT) context. Hereby, cars need to upload and transfer their data via a mobile communication network into the cloud for further evaluation. As wireless resources are limited and shared by all users, data transfers need to be conducted efficiently. Within the scope of this work three car-to-cloud data transmission algorithms Channel-Aware Transmission (CAT), predictive CAT (pCAT) and a periodic scheme are evaluated in an empirical setup. CAT leverages channel quality measurements to start data uploads preferably when the channel quality is good. CAT's extension pCAT uses past measurements in addition to estimate future channel conditions. For the empirical evaluation, a research vehicle was equipped with a measurement platform. On test drives along a reference route vehicle sensor data was collected and subsequently uploaded to a cloud server via a Long Term Evolution (LTE) network.