THIS IS A NOTICE OF INTENT TO AWARD A SOLE SOURCE CONTRACT AND IS NOT A REQUST FOR COMPETITIVE QUOTES. The US Army Corps of Engineers, Engineer Research and Development Center (ERDC), Environmental Laboratory (EL) intends to issue an award on a sole source basis (IAW FAR 13.106-1(b)(1)) to Bennett Aerospace, Inc. CAGE code: 61DA9 LOCATED 1 Glenwood Ave, Raleigh, NC 27603 for Model Development of Inhaled Toxicants and Neural Network Model Development and Optimization Task 1- Model Development of Inhaled Toxicants

The contractor shall merge a machine learning model developed to predict wind velocity fields in urban environments with a computational framework developed to assess the risks of toxic industrial chemicals (TICs) and toxic industrial materials (TIMs) released into the air at a proposed location. The contactor shall also integrate this modified risk model (mRM) with algorithms and models that deliver how the toxin moves through urban environments. The main goal from this research task is to develop a single computing platform that is able to incorporate exposure risk from high fidelity wind flows that guide toxic plumes around buildings in an operational area. Both the computational model and framework produced should integrate computer code that can be utilized across different programming languages and packages. Code should be suitable for both high performance computing platforms (e.g., HPC) and standard desktop computers. Lastly, the model shall be tested in defined test locations which allow for a vast array of expected outcomes.

Task 2 – Neural Network Model Development and Optimization

The contractor shall develop several biologically based neural network models from field data. The developed model should correlate input data such as aerial spectral data and drone collected images with desired outputs, such as field measurements of size/location/type of ground cover, aquatic plants, or understory layer of vegetation. Additionally, the neural network model should be able to train input data to some kind of regression output that approximates a statistical distribution. For example, input data should correlate with higher order moments of the data distributions collected for observations like foliage size, height, or spatial extent. Neural network models should be formulated with multiple layers and layers of different size.