HYPERKELP INC — Department of Defense SBIR Phase I: N231-013

Description

While Electro Optic (EO) sensors stationed on the surface and space systems are key to the Deprtment of the Navy’s (DoN) ability to track and counter hypersonic threats, they face limitations. Both space based and surface level sensors are prone to signal blockage by clouds, which absorb optical and infrared signals characteristic to Hypersonic Glide Vehicles HGVs. Well over 50% of the ocean is subject to cloud cover at a given time, meaning that the majority of deployed surface based EO sensors will face significant challenges in detecting and identifying obscured threat signatures. These obstacles demand any successful tracking system to take a multimodal approach to the detection of HGVs, especially during the glide and terminal phases of flight.  HyperKelp proposes that any effective counter hypersonic EO sensing mesh will incorporate tip-and-cue capabilities from rapidly deployable infrasound and acoustic sensors. By fusing EO signals with acoustic and infrasound signals, signatures of overhead hypersonic threats cannot be masked by cloud cover or changes in velocity. Such signals have been used to study hypersonic craft at large standoff distances since the Apollo missions, but only recently have advancements in edge computing enabled the classification of the source vehicles. HyperKelp’s autonomous sensor platforms will incorporate this technology to host acoustic sensors and novel, non-traditonal signal processing models to act as classifier engines. If successful, these automated classifiers will provide tip-offs for cooperative sensor array technology, including optical sensors further along threat axes. This multimodal approach is resilient to weather conditions, increases preparedness throughout the killchain, and offers redundant and independent sources of information that will strengthen all EO-based detection systems.  This Phase I will deliver a feasibility study of a deep machine learning technique to train Dense Neural Networks (DNN) as acoustic signal source classifiers. Building on HyperKelp’s existing body of work, these DNNs will analyze real time spectrogram feeds to identify acoustic signal structures, like N-wave overpressures, that are consistent with overhead hypersonic vehicles. This work will culminate in a plan to prototype this passive sensor and target classification capability onboard HyperKelp’s autonomous Kelp Smart Buoy (KSBM) platforms during a Phase II. The Phase I Option will integrate and deploy the Phase I products aboard HyperKelp’s existing KSBM system to accelerate the project toward delivery of a mission-ready product by the end of a Phase II.