With recent programs that enable the airborne launch and recovery of UAVs, there are opportunities to provide aerial threat assessment in difficult to deploy areas. A valuable enhancement to the airborne-deployed UAV system are sensors that detect invisible threats such as chemical, biological, radiological, nuclear, and explosive (CBRNE) weapons. The Intelligent Plume Mapping Payload, proposed by Deep Analytics LLC (DA), is a chemical and radionuclide sensing UAV payload that uses cutting edge machine learning techniques to estimate the plume source and extents in real time with the ability to automatically re-task the host UAV to sample additional regions of the plume as needed. DA has partnered with the University of Vermont to provide their cutting edge plume mapping approach to bolster DAs experience with payloads, UAV simulators, CBRNE sensors, and embedded computing. The phase I project focuses on adapting the machine learning method (a modified counter propagation neural network) to airborne plume mapping using simulated plumes, an assessment of the characteristics of chemical and radionuclide sensors for the UAV domain (e.g., effects of wind, prop wash, humidity, etc.), and the development of a hardware-in-the-loop simulator integrating the sensor hardware, our existing UAS flight simulator, and the simulated plume.UAV,machine learning,Plume Mapping,chemical sensor,RadioNuclide Sensor
