Modeling human supervisory control in heterogeneous unmanned vehicle systems

Abstract

Given advanced technology that relieves the human operator of low-level tasking and the future vision for network-centric operations, operator supervisory control of Unmanned Vehicle (UV) teams is likely to be a focal point of future research and development. Due to requirements for interoperability among UVs of varying attributes, heterogeneity in vehicle capabilities and tasks is likely to exist in future UV systems. This will lead to a large design space for these systems, which will cause design validation to require lengthy and expensive human-in-the-loop experimentation. This problem is addressed in this thesis through the following: First, identification of human-UV interaction attributes and associated variables that should be captured when modeling supervisory control of heterogeneous UV systems. Second, the derivation of a queuing-based multi-UV discrete event simulation (MUV-DES) model that captures both vehicle-team variables (including team composition and level of autonomy) and operator variables (including attention allocation strategies and situational awareness). The MUV-DES model supports design validation by simulating the impact of alternate designs on vehicle, operator, and system performance. To determine the accuracy and robustness of the MUV-DES model, an Internet-based test bed was developed to support extensive and rapid data collection for supervisory control of multiple heterogeneous UVs. Using data accumulated from online experiments, a multi-stage validation process was applied. The validation process resulted in achieving confidence in the model's accuracy and determination of the model's robustness under different input settings. Following the validation process, the MUV-DES model's ability to aid in the design and assessment of heterogeneous UV teams and related technologies was evaluated.

(cont.) More specifically, the MUV-DES model generated design recommendations addressing three underlying research objectives: a) indicating how potential operational/developmental design modifications could lead to performance improvements including 30% reductions in average vehicle wait times, b) identifying potential capabilities and limitations of future designs, including the detrimental impact of service time heterogeneity greater than 40% on average vehicle wait times, and c) replicating observed behavior in an existing system as a means of diagnosing the causes of vehicle-performance inefficiency. A subset of the MUV-DES model design recommendations was then implemented and the predicted benefit was validated using an additional set of experiments.