Energy-Aware Deep Learning for Green Cyber-Physical Systems

Abstract

Today's green computing has to deal with prevalent Cyber-Physical Systems (CPSs), engineered systems that tightly integrate computation and physical components. Green CPS aims to use electronic/computer devices and resources to perform operations as efficiently and eco-friendly as possible. With the rise of smart technology combining with Artificial Intelligence Deep Learning (DL) in Internet of Things and CPSs, continuing use of these compute intensive CPS software like DL can negatively impact energy resources and environments. Much research has advanced green hardware and physical component development. Our research aims to develop green CPSs by making them energy aware. To do this, we propose an analytical modelling approach to quantifying energy consumption of software artifacts in the CPS. The paper describes the approach through energy consumption modelling of DL in distributed CPS due to the popular deployment of DL in many modern CPSs. However, the approach is general and can be applied to any CPS. The paper illustrates the application of our approach for energy management in scaling and designing smart farming CPS that monitors crop health.