USV/UUV Autonomous Behavior Development-2

USV (Unmanned Surface Vehicle) and UUV (Unmanned Underwater Vehicle) missions generate large volumes of data that must be continuously evaluated. Automated pattern and anomaly recognition remains a daunting problem for USV/UUV missions. The objective of this program is to provide an automated system that will allow UUVs the ability to detect and avoid mines and other hazardous objects, with accuracy comparable to a modern manned surface or submersible craft. This problem encompasses all of the issues associated with auditability, opacity, and “big data” including storage, search, and filtering. Artificial intelligence and deep learning systems used for image analysis and pattern recognition have a significant limitation: there is no way to explain how their decisions are made. Conventional edge detection techniques result in too many false positives; this is because they are not cognizant of shape and have no awareness of the photometric characteristics of physical objects texture characterization or topographical features. No current commercial technologies exist that address the military applications that the Navy seeks. Vy has developed a powerful set of algorithms called Shape Based Modeling Segmentation (SBMS). We apply transparent and auditable mathematical models (Bézier curves and decision trees) to collect hard data from visual imagery that significantly increases the speed and accuracy of object recognition. In Phase I of this Project, we successfully demonstrated that Vy’s novel SBMS technology can be applied to detect mines and subsurface obstacles in an automated way that is compatible with Navy standards such as COIN (Common Operator Interface) and UMAA (Unmanned Maritime Autonomy Architecture). Our shape detection technology is well suited to the challenges of object detection and collision avoidance under harsh conditions where many sources of noise and distortion are present. An important benefit of Vy’s SBMS technology is that it provides a data representation architecture that fuses data from multiple sources (hyperspectral, multi-sensor and multi-mode) into an integrated format, assigning a numerical characterization of the credibility of each signal based on the degree to which that signal is corroborated by other sensor responses. All results of SBMS are saved in an industry standard database that is auditable and transparent. These capabilities have important implications for improving the reliability of decisions made by the UMAA autonomy stack in a UUV or USV.