Metrics
3 File views/ downloads
6 Record Views
Dissertation
A Predictive Model for Understanding and Enhancing Scalability of Human-Multi-Agent Teams
University of West Florida Libraries
Doctor of Philosophy (PHD), University of West Florida
2025
Abstract
The effective scalability of human-multi-agent teams is a critical challenge as autonomous systems grow more sophisticated, yet existing fan-out models for predicting supervisory performance are often inaccurate. Legacy models produce overly optimistic predictions because they are built on ambiguous variables that conflate robot capability with human behavior, and they are structurally limited to one-to-one interaction paradigms.This dissertation develops and empirically validates a new, comprehensive framework for modeling fan-out that addresses these flaws. The framework’s core contribution is the deconstruction of the ambiguous neglect time variable into two distinct components: a measure of pure robot capability and a measure of the human’s attentional strategy. This approach enables a more accurate, bounded model where potential fan-out defines a theoretical upper limit under ideal conditions, and actual fan-out is determined by the measured alignment between human and robot.
Across three successive experiments, the new model was shown to be a significantly more accurate predictor of measured fan-out than legacy models. Key findings demonstrate that: (1) scalability is driven by the alignment between robot capability and human attentional strategy, not robot autonomy alone; (2) the model’s bounded nature allows for robust performance and calibration improvements where unbounded legacy models fail; and (3) group-based interaction
schemes, while necessary for high-scale performance, introduce a quantifiable trade-off between team scale and the operator’s observability and directability over individual agents. The extended model accurately captures performance in both traditional one-to-one and modern group-based scenarios that render legacy models obsolete.
Ultimately, this work provides both a new conceptual lens and a validated mathematical tool for analyzing human-multi-agent team scalability. By accurately modeling the dynamic interplay between human and machine and providing the first predictive model for group-based interactions, this framework offers a more robust foundation for designing and deploying the effective human-autonomy teams of the future.
Details
- Title
- A Predictive Model for Understanding and Enhancing Scalability of Human-Multi-Agent Teams
- Resource Type
- Dissertation
- Contributors
- Hakki Erhan Sevil (Committee Co-Chair)Matthew Johnson (Committee Co-Chair)K. Brent Venable (Committee Member)Jeffery Phillips (Committee Member)Niranjan Suri (Committee Member)
- Publisher
- University of West Florida Libraries; Argo Scholar Commons
- Format
- pdf
- Number of pages
- 175
- Copyright
- Permission granted to the University of West Florida Libraries by the author to digitize and/or display this information for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires the permission of the copyright holder.
- Identifiers
- 99381563233106600
- Academic Unit
- Intelligent Systems and Robotics
- Language
- English
- Awarding Institution
- University of West Florida; Doctor of Philosophy (PHD)
- Theses and Dissertations
- Doctor of Philosophy (PHD), University of West Florida