The project is a continuation of the project from Year 6 with the identical title.
In Year 6, through the project, a modeling and solution framework based on hierarchical decomposition and reinforcement learning (RL) adaptive multi-scale prognostic and health management (AM-PHM) was developed. The framwork was able to provide a decision making policy for the maintenance and operation of an industry 4.0 industrial internet of things (IIoT) capable smart manufacturing facility.
In Year 7, the project is extended to focus on the scalability and implementation of the developed framework. The research activities will be centered around a) development of scalable smart manufacturing simulation with IIoT capability, b) comparison study of the scalability of AM-PHM against deep RL methods such as asynchronous advantage actor-critic (A3C) methods, and c) apply and improve RL methods (based upon Wolpertinger etc.) on manufacturing settings, and see how the concept of resource allocation is realized via deep RL, specifically under the settings with large-scale discrete action space.