The investigator plans a three year research program to develop algorithms for sensor systems for the detection of chemical and biological materials. This work builds on prior research of the investigator and her colleagues involving autonomous mobile sensors for environmental sampling and algorithms for understanding hyperspectral imagery data. The research program involves the design of multiscale, multimodal sensing and detection algorithms, using data from both standoff detection and point detection from sensors mounted on mobile autonomous platforms. This data-intensive research depends on the modes of data available and their spatio-temporal resolution, viewpoints, and spectral resolution. The work includes the design and construction of a numerical simulator for the project, that incorporates various sensing modalities and on which algorithms are tested against against field data supplied by the government. In addition, mobile sensing algorithms are validated and tested at a laboratory multi-vehicle wireless testbed involving simpler sensors as a proxy for field sensor data. The research exploits recent algorithmic advances in image analysis and reconstruction from high dimensional data. These include, but are not limited to, compressive sensing methods, total variation minimization methods, hybrid wavelet-PDE algorithms for data fusion at different scales, hybrid geometric-stochastic algorithms for real time path planning and analysis, and nonlinear filtering.
The ability to detect and analyze biological and chemical threats in real time is essential to the future security of our country. Recent advances in sensor design now allow for rapid collection of information from multiple vantage points, involving multispectral sensing modalities. Where we are lacking is the ability to rapidly process and understand evolving information from diverse platforms to accurately identify and track the threat. This challenging problem requires new ideas for mathematical algorithm design to fuse the diverse data and provide accurate detection with both a low false alarm rate and detection delay. This research program develops new methods for high performance data processing and new fast algorithms for identification, in order to optimally utilize state-of-the-art and future sensor technology.
