Digital-Twin-Aided Product Design Framework For IoT Platforms
Chenyu Wang, Yingshu Li
Abstract
The increasing number of products is the trend of current industry. However, the product development process is significantly limited by budget and testing risk. Recently, digital twin (DT) has emerged as a promising industrial paradigm that provides an integrated and cohesive view of the product design process. In this article, we propose a product design framework for Internet of Things (IoT) platforms, namely, DT-aided IoT platform design (DTIPD). This framework considers a large number of IoT devices performing different tasks with machine learning (ML) technologies. Each IoT device constructs a particular ML-based model that deals with its task automatically by feeding related data with labels. The challenges of large-scale network management and ground-truth shortage at the initial stage of product iteration are addressed. We propose a two-level hierarchical learning process using the real-time model status stored at DT servers (DTS), aiming to improve product quality while shortening the development lifecycle. The comprehensive experimental results for both the single-DTS and multiple-DTS scenarios demonstrate the applicability of our framework.