BRITE Fellow: Autonomous Systems that Accommodate Human Perception and Reasoning about Uncertainty

点击翻译按钮获取中文摘要

This Boosting Research Ideas for Transformative and Equitable Advances in Engineering (BRITE) Fellow award will fund research that enables predictably safe interaction between autonomous systems and humans in uncertain environments, with applications in aerospace, manufacturing, transportation, and healthcare systems, thereby promoting the progress of science and advancing the national prosperity, welfare, and defense. By sensing and reacting to their environment, autonomous systems have the potential to help human operators accomplish difficult, dangerous, or risky tasks more easily and safely. However, when autonomous systems appear to be unpredictable, unreliable, or unresponsive, they become hindrances, rather than a source of help. To prevent this outcome, this project aims to develop an algorithmic design and control framework for autonomous systems that successfully accommodates the inevitable unpredictability of human actions, as well as the effects on human decision-making of uncertainty in the dynamics and action of the system. A key element of this framework is its ability to capture typical needs of humans to design solutions from uncertain information, and under potentially conflicting constraints, which meet desired objectives to the greatest possible degree. This research is integrated with curricular initiatives aiming to promote culturally responsive pedagogy and project-based learning in the training of students from diverse backgrounds in human-centric design of autonomous systems.This research aims to make fundamental contributions to a stochasticity-based methodological approach for integrating knowledge of human perception and reasoning about uncertainty into the design and control of autonomous dynamical systems. To this end, new mathematical theory and computational algorithms will be developed, based in control theory, machine learning, and human factors. Such theory will address the handling of arbitrary, non-Gaussian forms of stochasticity and rational and non-rational decision models, with emphasis on computationally efficient controller synthesis. The feasibility and mathematical properties of methods to constructively accommodate human variability, and to be context-aware, will also be explored. The developed algorithms and theories will be experimentally validated in simulation-based platforms. This project enables the principal investigator to leverage expertise in methods for assuring probabilistic safety in stochastic dynamical systems and ongoing work on integrating cognitive models in autonomous systems to pursue a high-risk vision for human-centric autonomy with significant potential for transformational impact. It lays a foundation for advancing culturally responsive teaching and research practices throughout the engineering community.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.