INTERACTIVE MACHINE LEARNING BASED ON DEEP REINFORCEMENT LEARNING AND GENERATIVE ADVERSARIAL NETWORK HYBRID FOR DIGITAL TWIN

Abstract

A digital twin (DT) system should respond to unpredictable changes quickly which requires planning to be performed based on the real time operating conditions and dynamic changes to be handled with cognitive skills. Majority of existing works focus on fully autonomous ability but having human in the decision making loop is still critical especially in problems that are dynamic, inherit high complexity and the rules cannot be expressed explicitly. Existing learning approaches provide reliable and efficient real-time control and management of the system elements based on deep reinforcement learning (DRL) which requires huge training size. The performance of this approach depends on the distribution of the data in the learning environment. Therefore, this study proposes an interactive machine learning (iML) approach based on the hybrid between DRL and generative adversarial network (GAN) for supporting human’s asset maintenance planning. The focus of the study is to design adaptive agents that support meaningful and beneficial interaction with humans. The iML approach will consist of two functions: (a) the DT will collect data about the asset to allow humans to monitor the asset’s health, (b) the DT will then semi-autonomously learn the human’s intervention decision making using DRL and GAN. The goal of DRL and GAN agent in the DT is to maximize a reward function in optimising the planning and scheduling whilst including human interaction for semi-autonomous decision making. Dataset for asset health monitoring and prognostics will be used for demonstrating the performance of the DT. The performance of the iML through DRL+GAN will be compared against these approaches individually.

Document Details

Document Type
DoD Grant Award
Publication Date
Nov 16, 2022
Source ID
FA23862114050

Entities

People

  • Sharef, N. M.

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • University of Putra Malaysia

Tags

Fields of Study

  • Computer science

Readers

  • Fault Tolerant Diagnosis of Black and White Balloon Isolation Tests Using ¥.
  • Neural Network Machine Learning.
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Neural Networks