EECS PhD student Muhammed Golec, under the supervision of Professor Steve Uhlig and Dr Sukhpal Singh Gill, leads the CAPTAIN Testbed, a research project that pioneers the integration of Artificial Intelligence, Internet of Things, and serverless computing to streamline predictive maintenance in industrial settings, which is published in the IEEE Internet of Things Journal.
The IIoT and AIoT-driven manufacturing generate enormous volumes of data, necessitating highly scalable, energy-efficient systems. Traditional cloud setups often lead to high latency and energy consumption, contributing to higher CO2 emissions. CAPTAIN’s serverless computing framework dynamically adjusts resources to user demands, enabling efficient data processing and predictive maintenance. This adaptability is crucial as IIoT applications scale, often handling unpredictable data surges. CAPTAIN's AI models, particularly the Random Forest (RF) and Light Gradient-Boosting Machine (LightGBM), address "cold start" latency challenges typical of serverless systems, where resources are allocated on-demand, thus optimizing response times and energy usage.
Within the CAPTAIN framework, IIoT devices capture operational data on variables such as temperature and rotational speed, continuously monitored and analyzed by the AI module to predict and prevent system failures. When cold start latency is predicted, CAPTAIN’s AI-driven "ping" mechanism initiates resources in advance, maintaining system responsiveness. In tests using Google's serverless environment, CAPTAIN outperformed baseline models, achieving significant improvements in latency and scalability. The CAPTAIN framework demonstrates that serverless platforms, enhanced by AI models, can achieve the required scalability and efficiency for IIoT applications, particularly in critical predictive maintenance scenarios.
While CAPTAIN already minimizes downtime and maintenance costs, upcoming enhancements will aim to integrate security protocols, such as Blockchain and OAuth 2.0, to secure data transfers within the serverless architecture. By exploring AI-based load-sharing methods across cloud and edge nodes, future iterations of CAPTAIN will broaden its applicability to more time-sensitive tasks within Industry 4.0, potentially expanding to areas like autonomous industrial systems and real-time health monitoring.
For those interested in advanced industrial solutions, the CAPTAIN framework exemplifies how serverless computing, integrated with AI and IoT, is shaping the future of predictive maintenance and intelligent system management in Industry 4.0.The CAPTAIN Testbed is released as open-source software to advance the research. The implementation code, datasets, result reproducibility scripts are publicly available, can be found at the GitHub repository.
A YouTube video explains this work, enabling future authors to easily understand its workings.
Publication Details
Muhammed Golec, Huaming Wu, Ridvan Ozturac, Ajith Kumar Parlikad, Felix Cuadrado, Sukhpal Singh Gill and Steve Uhlig, "CAPTAIN: A Testbed for Co-Simulation of Scalable Serverless Computing Environments for AIoT Enabled Predictive Maintenance in Industry 4.0," in IEEE Internet of Things Journal, 2024. doi: https://doi.org/10.1109/JIOT.2024.3488283
Read the full publication.