Edge AI-Empowered Smart Devices and Robotics for AIoT Applications
Principle Investigator :
Prof. Jiannong Cao
Dean of Graduate School, Otto Poon Charitable Foundation Professor in Data Science, Chair Professor of Distributed and Mobile Computing, Director of RIAIoT, Director of UBDA
Edge AI is an innovative technology that combines edge computing and artificial intelligence to enable real-time data processing and intelligent decision-making on IoT devices and robots. We have developed an edge AI platform that supports faster and collaborative model training and inference. Specifically, we have developed an edge-native task scheduling system to manage large-scale, geographically distributed, and heterogeneous edge resources. Atop this system, we have designed various resource-aware scheduling algorithms to optimise AI model training and inference, taking into account both AI model characteristics and underlying resources. Additionally, we offer easy-to-use programming APIs to streamline the development of edge-native AI applications.
To further demonstrate the advantages of edge AI, we have also designed and developed an edge AI robot to inspect pipelines. This robot stands out due to its three key features: its real-time AI-based defect detection, its deformable design, and its autonomous control. It uses advanced edge AI technology to enable compressed, optimised AI models, enabling it to detect pipeline defects in real time. This feature makes the robot effective in challenging environments, such as underground or underwater pipelines. The robot’s unique deformable design and self-control algorithm also enables it to adapt and navigate through various pipe structures.