Aerial Platforms Artificial Intelligence at the Edge (AP-AI)

TECHNICAL ABSTRACT The proposed system will be called the Aerial Platform Artificial Intelligence at the Edge (AP-AI). The goal of this project will be set to cover the widest range of aerial platforms in the AFSOC inventory (AC-130, MQ-9, U-28, CV-22, Armed Overwatch) with specific tuning of the outcome of this effort to experimentation involving the AC-130J. We anticipate significant benefits in the Defense Mission Need, especially to Moving Target Engagement, Tactical Air Dominance, and Resilient Basing. The technology produced will have significant components that will address Army and Navy needs with little or no modification. The data collected by EO/IR and other sensors mounted on the AC-130J gunship are often degraded by factors such as noise, low contrast, jitter, and atmospheric aberrations. The native resolution and frame rate of these sensors are not adequate to detect and track small and high moving targets. We anticipate significant benefits in the Defense Mission Need, especially to Moving Target Engagement, Tactical Air Dominance, and Resilient Basing. To remedy these inadequacies, we place a processing unit between the sensors and display manned by an operator to:Enhance video streams.Increase the resolution of the video collected.Automatically detect threat targets.Achieve optimal frames per second data collection.Run full system with minimum size, weight, and power. The main objective of this SBIR is to implement and validate the 5 improvement areas listed above by deploying the processing system on the AC-130J platform. Because the processing system is sensor-agnostic; it applies to many other sensors mounted on ground and surface platforms.  The base technology developed by the firm enhances video, super resolves it, and can extract ID and location data of objects. The current level of performance indicates that this technology outperforms most of competing systems. Tests conducted by the Emerging Technology Laboratory on a C-208 Caravan equipped with MX-15 and MX-20 turret ball sensors (the same sensor package mounted on the AC-130J) for side-by-side comparison have shown significant improvements. Phelps2020 has the tools and a detailed plan to:Enhance further the video.Super resolve the video to higher level of magnification.Train with a larger data set including synthetic data sets.Deploy the system on an aerial platform on the ground and in the air.Determine the envelop of performance using standard metrics. Insert the processor between sensor and display manned by operator.Perform machine-to-machine communication for automation with or without a man-in-the-loop. Phelps2020 will combine computer technologies such CPU, GPU, and/or FPGA to reduce size, weight, and power (SWaP) to ensure accuracy of the results and real-time performance. All the implementations will be benchmarked against established metrics in the field, which is one of the biggest features that the AFSOC customer and end-user liked in this work plan.