Techniques for improving localization applications running on low-cost IoT devices

Nowadays, localization features are widespread in low-cost and low-power IoT applications such as bike-sharing, off-road vehicle fleet management, and theft prevention of smart devices. For such use cases, since the item to be tracked is inexpensive, older or power-constrained (e.g. battery-powered vehicles), localization features are realized by the installation of low-cost and low-power devices. In this paper, we describe a set of low-computational power techniques, targeting low-cost IoT devices, to process GPS and INS data for accomplishing specific and accurate localization and tracking tasks. The methods here proposed address the calibration of low-cost INS comprised of accelerometer and gyroscope without the aid of external sensors, correction of GPS drift when the target position is static, and the minimization of localization error at device boot. The performances of the proposed methods are then evaluated on several datasets acquired on the field and representing real use-case scenarios.