Autonomous Coordination Policies in Ground-Air Unmanned System Interaction
Abstract
Humans use a variety of complex strategies when solving problems. These strategies are semantically ordered; that is, as the complexity of a problem increases, an explanation of a strategy used for a simpler problem would be related to that of a strategy used for a more complex problem. Machines, however, can learn complex strategies but they normally produce strategies that are different from those used by humans. This leads to loss of trust and ineffective human-machine teaming. We designed machine learning algorithms that can learn incrementally while preserving the meanings embedded in a simpler context as they learn a more complex one. One proposed apprenticeship bootstrapping algorithm has been tested in simulation and physical environments on ground-air interaction tasks, while the others were either demonstrated theoretically or tested on simulated synthetic tasks alone. The research has demonstrated that machine education and bootstrapping techniques should be integral parts of explainable and interpretable machine learning to assure a trustworthy learning machine.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 12, 2023
- Accession Number
- AD1226616
Entities
People
- Hussein A. Abbass
Organizations
- University of New South Wales