š¬ Research Summary by Wander Siemers, MSc student at Delft University of Technology with an interest in mobile technology and sustainability.
[Original paper by Wander Siemers, June Sallou, and LuĆs Cruz]
Overview: Artificial intelligence (AI) is both a key enabler of desired mobile functionalities and a major power draw on these devices. In this paper, we present a review of the literature of the past decade on the usage of artificial intelligence within the realm of green mobile computing, taking both aspects of mobile artificial intelligence into account.
Introduction
Information technology will use up to 21% of global electricity production in 2030. The smartphone market, in particular, has grown rapidly to over 7 billion devices in 2023. Smartphone battery life has not grown in tandem and is still a major concern, especially with new energy-intensive applications like on-device machine learning. This research aims to analyze the literature surrounding Green Mobile AI using a Systematic Literature Review to increase our understanding of this research field. Thirty-four relevant papers were identified, and their topics, roles of AI, industry involvement, study level, and tool availability were analyzed.
Key Insights
AI is bringing ever-new functionalities to the realm of mobile devices that are now considered essential (e.g., cameras, voice assistants, and recommender systems). Yet, operating AI takes up a substantial amount of energy. However, AI is also being used to enable more energy-efficient solutions for mobile systems.
Hence, it has two faces in that regard, as both a key enabler of desired (efficient) mobile functionalities and a major power draw on these devices, playing a part in both the solution and the problem. In this paper, we present a review of the literature of the past decade on the usage of artificial intelligence within the realm of green mobile computing. From the analysis of 34 papers, we highlight the emerging patterns and map the field into 13 main topics that are summarized in detail.
Results
Our results showcase that research interest in the field has been slowly increasing in the past years, specifically since 2019, with an increase in the number of published papers. Regarding the double impact AI has on mobile energy consumption, the energy consumption of AI-based mobile systems is under-studied compared to the usage of AI for energy-efficient mobile computing, and we argue for more exploratory studies in that direction. The topics of the studies are often highly specific to a relatively narrow domain, such as federated learning. Although most studies are framed as solution papers (94%), the majority do not make those solutions publicly available to the community. Moreover, we also show that most contributions are purely academic (28 out of 34 papers) and that we need to promote the involvement of the mobile software industry in this field. Lastly, having access to up-to-date benchmarks is a major challenge in this field.
Between the lines
We describe two main branches in the literature: 1) papers looking at the energy consumption of mobile AI applications and 2) papers focusing on applying AI to reduce mobile energy consumption. We identify groups of papers that consider similar topics or use similar techniques. We pinpoint main research directions, such as offloading and networking optimization, to save energy on mobile devices and analyze the energy consumption of federated learning. However, other areas, such as approximation computing, have been investigated less intensely.
For researchers, this paper provides an overview of this research area, and it points to promising directions for future research. It is also relevant for stakeholders in the mobile computing industry as we identify potential solutions that arise from deploying AI models in mobile apps. It also helps identify areas where further research and investment are needed, such as the lack of industry involvement, low availability of tools, and lack of observational studies on the subject.