Maintaining Synchrony of Dual Machine Learning: A Phase-Locked Loop Approach

Abstract

Smart home systems have shown potential to enable older adults to age-in-place, delaying entry to care. However, previous work has revealed network inefficiencies in these systems. For telecom carriers, these findings become more significant with the wide-scale deployment of smart home systems and, more generally, Wireless Sensor Networks (WSNs). Subsequently, research applied Dual Machine Learning to reduce network traffic leaving the residence to cloud processing. However, the dual model was shown to be impacted by network effects such as latency, jitter, and packet loss, whereby as much as half of sensor data stored in the cloud was incorrect. This report proposes a 2-stage Phase-Locked Loop (PLL) based solution to mitigate the impact of network latency and jitter on Dual Machine Learning and improve the accuracy of data stored in the cloud; the proposed solution increased the worst-case accuracy rate from 71.4% to 94.6% for latency and from 64.1% to 90.3% for jitter.