By providing a unified approach to modeling & simulation for training that leverages deep domain knowledge of acoustic modeling within a state-of-the-art machine-learning and knowledge-engineering framework, ARiA is developing training systems that enhance the effectiveness of our nation’s warfighters. Our current training projects include

Environment for Surface ASW Interactive Learning (ESAIL™)

To address the ASW training challenge of teaching Navy sonar operators advanced concepts and tactical-system operation in complex environments given constraints on training and compressed training cycles, ARiA is developing ESAIL™, the Environment for Surface ASW Interactive Learning. Building on the experience developed during the creation of WaveQuest™, ESAIL™ couples intuitive three-dimensional visualization of the ground-truth tactical situation and environmental data with real-time interactive rendering of recorded and simulated sonar data to help operators gain a deep understanding of (1) the relationships between observed phenomena on sonar tactical displays and the corresponding phenomena and environmental conditions in the physical world and (2) how an understanding of the mapping between the two domains can be used to improve tactical-system employment.


In partnership with award-winning game company Filament Games, ARiA is developing a capability for interactive game-based education in underwater acoustics that incorporates real-time fidelity-adaptive modeling & simulation to realize both STEM education goals for secondary students and training goals for US Navy sonar operators. Together ARiA and Filament are producing WaveQuest™, which will be commercially available to students, parents, and educators.

Synthetic ASW Generation Engine (SAGE™)

The Synthetic ASW Generation Engine (SAGE™) is a fundamentally new approach to automatic scenario generation for simulation-based training. Under development for the ONR 321US High-Fidelity Active Sonar Training (HiFAST) EC effort, SAGE integrates an inference engine and ontology of ASW sonar training within an active semisupervised machine-learning framework to enable automatic generation of scenario parameters for fidelity-adaptive simulation-based ASW sonar training that jointly satisfy training objectives and computational-load constraints.
Effective signal processing in complex environments requires an understanding of the physical processes that generate noise and clutter in those environments.

By leveraging our work in modeling & simulation, ARiA is developing signal-processing techniques and algorithms that enable automation of detection, classification, and recognition that is robust in the most challenging environments. Our current signal-processing projects include

Deep Representation Learning from Sonar Big Data for Automatic Target Recognition

ARiA is leading a team including scientists from The Pennsylvania State University Applied Research Laboratory (ARL/PSU) investigating the application of recent deep-learning approaches for unsupervised feature-space learning to large sets of data gathered by operational mine-countermeasures (MCM) and explosive ordnance disposal (EOD) sonar systems. Deep learning has resulted in significant gains in performance of commercial image-recognition systems by learning semantically meaningful feature representations from unlabeled data. Our current work is extending these advances to Navy applications.

Topological Investigation of Target/Clutter Features in Mine-Countermeasures Sonar Data

ARiA is leading a team including mathematicians from American University to leverage recent developments in topological signal processing including algebraic-geometric methods and stratified-manifold learning toward development of new and insightful feature-space representations of scattering from targets and clutter. In addition to development of new feature representations for automatic target recognition (ATR), this effort is investigating the use of topological methods to characterize computational models used for simulation of mine-countermeasures (MCM) sonar systems, such as PC SWAT.

Mitigation of Biologically Induced Active Sonar Reverberation in Littoral Regions

ARiA, with support from Applied Research Laboratories – The University of Texas at Austin (ARL:UT) is formulating and developing new concepts and algorithms for data-driven clutter-adaptive waveform-synthesis and CFAR normalization processing based on physical models of resonant backscattering from heterogeneous aggregations of swim-bladder-bearing fish that will significantly improve active sonar detection capability in littoral waters by reducing the number of false contacts and decreasing the amount of display clutter.