Abstract
IoT systems are used for performing a variety of essential tasks in wide-ranging sectors, such as healthcare, smart cities, agriculture, and industrial automation. Most of these systems incorporate smartness for intelligent decision making using artificial intelligence (AI) approaches. To reduce the network bandwidth usage and load on the cloud server, there is a push to relegate most of these AI-related computations to edge IoT devices/systems. Hence, to ensure good performance, the edge devices must be systematically designed with an emphasis on the respective AI requirements from exploration to deployment. State-of-the-art IoT device and system design practices place little importance on the AI specifications during the early stages of the system development resulting in systems, which are unable to meet the AI specifications without additional redesigning and optimization efforts. In this letter, we propose an automated framework for AI-rooted IoT system design approach, where the AI specifications play a vital role in deciding the system components, design, and implementation from a very early stage of the design life cycle. The proposed framework employs an expert system and a machine-readable knowledge-base to automate the design process. Using a set of case studies, we demonstrate the benefits of the proposed framework.
Reiner Dizon-Paradis
Postdoctoral Research Associate
My research interests include machine learning applications in national security, hardware security and assurance, artificial intelligence of Things, and robotics.