KNOWLEDGE BASED SYSTEMS INC — Department of Defense SBIR Phase I: N193-A02

Description

KBSI will establish an Intelligent Anomaly-Based Maritime Situational Awareness System (IMSAS) for enabling intelligent self-contained acoustic detection capability within Unmanned Surface Vehicle (USV) and/or an Unmanned Undersea Vehicle (UUV). The concept is to build a framework for multi-dimensional anomaly detection, where multiple predictive models are trained to model the normative behavior for a taxonomy of pattern types. Advances in deep learning models would be leveraged for calibrating models that are specifically geared for the specific type of pattern and learn the extremely non-linear patterns from noisy data and then use these pattern models to detect indicate unusual situations. The collective capability of these anomaly models will enable optimal autonomous behaviors that can effectively respond to surface ships and submarines. The Phase I project will research, design, and demonstrate the technical feasibility of the IMSAS solution approach. Phase II will refine, enhance, validate, and mature the IMSAS solution for focused sea-based autonomous applications, leading to rapid technology transition and commercialization. Important innovations include (i) taxonomy of pattern types that collectively establish normative behavior, (ii) deep learning, LSTM, and probabilistic graph models that are specifically calibrated for specific patterns, (iii) formulations that enable dynamic and actionable situational awareness for autonomous behaviors.