UBER-HEB: Universal Biologically-inspired Environment for Research: Hierarchical Ersatz Brain

In Phase I of UBER-HEB, the Aptima team made significant progress designing a non-von Neumann hierarchical universal learning system based on the Ersatz Brain Architecture (EBA). An experiment using an EBA implementation showed emergence of module assemblies, a critical step in validation. In Phase II, we propose to focus sharply on an important, well-characterized problem ideally suited to the EBA: torpedo detection, classification and localization. Trained human operators are very good at determining if a set of signals in the marine acoustic environment come from a torpedo or not, while computer algorithms, despite decades of advances in signal processing and the vast increases in computer power, are not. Alion Sciences, a key member of the Aptima team, is a lead contractor on the Navy’s Anti-Torpedo Torpedo Defense System (ATTDS); we have been working with the Alion ATTDS PM to ensure that our Phase II work is closely aligned with ATTDS both technically and in terms of schedule. We will use realistic marine acoustic data from the beginning of Phase II to develop our system and to benchmark our progress. This clear focus will both help to advance the EBA in general and contribute to addressing a significant operational requirement.