Creating an Objective Methodology for Human-Robot Team Configuration Selection
Singer, Sharon Michelle
Akin, David L
MetadataПоказать полную информацию
As technology has been advancing and designers have been looking to future applications, it has become increasingly evident that robotic technology can be used to supplement, augment, and improve human performance of tasks. Team members can be combined in various combinations to better utilize their capabilities and skills to create more efficient and diversified operational teams. A primary obstacle to integrating new robotic technology has been the inability to quantitatively compare overall team performance between very different team configurations without limiting the analysis to a few metrics. To-date, mission designers have arbitrarily assigned importance to mission parameters, subjectively limiting the search space. While this has been effective at evaluating individual mission plans, the arbitrary evaluation criteria has made a straightforward comparison between different research projects and ranking scales impossible. The question then becomes how to select an objective set of criteria for any given problem. It is this final question that this research sought to answer. A methodology was developed to facilitate performance comparison amongst heterogeneous human and robot teams. This methodology makes no assumptions about mission priorities or preferences. Instead, it provides an objective, generic, quantitative method to reduce the complexity of the mission designer's decision space. It employs an heuristic, greedy objective reduction algorithm to reduce problem complexity and a multi-objective genetic algorithm to explore the design space. The human-robot team configuration selection problem was utilized as the application that motivated this research. The methodology, however, will be applicable to a wider domain of research. It will provide a structure to enable broader search of the design space, exploration of the differences between performance metrics, and comparison of optimization models that facilitate evaluation of the design options.