Abstract
The proliferation of low-cost and low-power IoT devices are constantly generating gigabytes data at the network edge. Bridging AI with IoT is a natural option to unleash the data on devices. AI-powered IoT systems can boost many novel applications and services with impressive performance. In this paper, we identify four key challenges for developing AI-powered IoT systems, including interpretability, efficiency, heterogeneity, and trustworthiness. We also summarize our proposed solutions to tackle each of these challenges. Finally, we discusses the open challenges and future research directions on designing inter-pretable, efficient, heterogeneous, and trustworthy IoT systems.
DOI
10.1109/CogMI52975.2021.00039
Year