Trustworthy and Sustainable Edge AI: A Research Agenda

Also stored on: https://pure.tudelft.nl/ws/portalfiles/portal/125931435/Trustworthy_and_Sustainable_Edge_AI_A_Research_Agenda_Taverne.pdf

Abstract

As a fast evolving domain that merges edge computing, data analytics and AI/ML, commonly referred as Edge AI, the community of Edge AI is establishing and gradually finds its way to connect with mainstream research communities of distributed systems, IoT, and embedded machine learning. Meanwhile, despite of its well-claimed potential to transform cloud and IoT industry, Edge AI is still a complex subject that faces critical challenges from the trustworthy and sustainable concerns. To shed light on these pressing matters, this paper aims to develop a research agenda for trustworthy and sustainable Edge AI. We clarify the concepts, define the proper scoping and propose a research agenda for Edge AI to be trustworthy and sustainable. To illustrate the research agenda in practice, we highlight two active R&D projects: the SPATIAL project on trustworthy Edge AI and the APROPOS project on sustainable computing. The projects serve as concrete use cases to explore the agenda development. Our goal is to equip researchers, engineers, service providers, government and public sectors with a better understanding of the underlying concepts and to raise awareness of emerging directions in trustworthy and sustainable Edge AI.