Project Grant 2153171

This Project Grant from the National Science Foundation's Computer and Information Science and Engineering program totaling $174,952 will support the development of explainable mathematical frameworks and algorithms for integrating multi-source sensor data from multiple modalities. The awardee, the Regents of the University of Michigan, will receive the funding from May 1, 2022 to April 30, 2024.

The project aims to advance fundamental knowledge in explainable data integration through a weakly supervised learning-based framework that accounts for both data and label uncertainties when fusing heterogeneous sensor datasets. Specific objectives include developing a scalable computational model leveraging binary fuzzy measures under multiple instance learning; incorporating bi-capacities and bi-polar Choquet integrals to model conflicts between data sources; and extending the existing multiple instance Choquet integral framework to sequential spatiotemporal data integration. Outcomes of the research are expected to facilitate decision-making and knowledge discovery from large volumes of multi-modal data across application domains involving remote sensing, robotics, agriculture and more.