This thesis contains three papers related to modeling a Human in the Loop (HITL) operator interacting with a system of autonomous vehicles in Intelligence, Reconnaissance and Surveillance (ISR) type Scenarios. The thesis begins with an empirical study of one such actual system, moves to a narrow-scope detailed experimental software simulation introducing probabilistic models and culminates in experimental investigation using a time dependent information tracking function with a set of analytically tractable probabilistic models. This first paper, entitled, ?Experimental Study of Information Load on Operators of Semi-Autonomous Systems? presents a set of experimental results studying the relationship between Human in the Loop performance and user workload in the RoboFlag test-bed. Operators played a series of games to evaluate performance as a function of information load (speed and number of vehicles). Results showed a positive relationship between game speed and total score. In addition operators reported using more automation as number of robots increased but trusting automation less as game speed increased.
The second paper, entitled ?Modeling Tradeoffs in Decisions by Operators Controlling Autonomous Vehicles? presents the results of two decision-making experiments and two operator decision models for an Intelligence, Surveillance and Reconnaissance (ISR) type mission. The first experiment used 25 possible scenarios, each of which included enough trials to allow a formal statistical model to be derived. The distribution of operator decision data was modeled with a binomial distribution as a function of environmental variables. An optimal decision-making policy was also prescribed for all scenarios. Results show good agreement between operator data and the optimal decision-making policy in most scenarios, except when the relative utility between the choices was similar. Lower order probabilistic models using conditional probabilities and Gaussian random variables are also derived; results show a strong ability to use lower order models for operator decisions. The second experiment presented operators with the same binary decision, but from a more general choice of 90 possible scenarios. This allows the evaluation of the probabilistic model as data becomes sparse. Operator data from the second experiment was successfully binned and compared to the results of first experiment, demonstrating consistent operator decision-making between experiments. The final paper, entitled ?Operator Decision Modeling for Intelligence, Surveillance and Reconnaissance Type Scenarios with a Time Dependent Information Function? presents a model of operator + vehicle interaction for a simplified Intelligence, Surveillance and Reconnaissance type mission utilizing a time dependent information function for target identification. The model is developed and evaluated using operator decision-making experiments where an operator controls a friendly uninhabited aerial vehicle (UAV) tasked with identifying enemy targets within a two-dimensional map. Operators must make two decisions: 1) which target to choose first, and 2) if and when to task the UAV to the second target to start data collection. Two sets of experimental data were collected. In all experimental scenarios, target choice and time on target were recorded. The data was analyzed in order to develop an analytically tractable model of operator choice. An optimal decision-making policy was also prescribed for all scenarios and compared to the operator data. Finally a both tabular and lower order probabilistic model developed to model decision making in this experiment.