SBIR-STTR Award

The U.S. Naval Services require a rapid force augmentation to meet growing threats from a variety of adversaries. Unmanned assets such as UUVs are the only feasible approach for meeting the force augmentation demands in a reasonable budget and timeline. In order to provide operationally useful capabilities, UUVs must be autonomous due to severe communication restrictions limiting human on- or in-the-loop control. As an additional communication challenge, the commander must have operational visibility over underwater autonomous assets such that she understands the mission state, even in intermittent communication environments. Digital twins are a modeling technology that, once adjusted for the underwater environment, can be augmented with machine learning and autonomy technologies to enable a variety of capabilities across a variety of missions relevant to the U.S. Navy. We propose DNA-mRNA (Digital twin Novel Architecture study with Machine learning Research for Naval Applications), a study to identify how digital twins can enable critical capabilities for the U.S. Navy, and how these capabilities can be operationalized on autonomous undersea vehicles. We will demonstrate that digital twins enable autonomous undersea systems to provide a variety of mission effects across a variety of missions, that they enable a commander to control, predict, and monitor an undersea system with communication constraints, and that they enable system adaptation to varying communication environments.

In order to maintain U.S. military advantages, the U.S. Naval Service must have widespread adoption of autonomous assets such as autonomous UUVs. However, autonomous UUVs present a variety of challenges in the maritime environment, such as (1) a requirement that the commander can predict, control, and understand an autonomous fleet, even under limited communications, (2) autonomous systems must be adaptable and resilient to changing platform characteristics, e.g., platform damage/degradation, and (3) autonomous systems at scale must minimize communication requirements while still coordinating with other platforms at the edge and ensuring the commander has current situational awareness of the fleet. We propose DNA-mRNA (Digital twin Novel Architecture Study with Machine Learning Research for Naval Applications). Under the Phase I effort, we demonstrated that digital twins augmented with Machine Learning and Autonomy techniques could effectively reduce communication requirements in a multi-UUV coordinating system, improve system resiliency and adaptability to unexpected events such as platform damage or degradation, and improved predictability and situational awareness for a commander. In the proposed Phase II effort, we will build a full prototype of this system concept, ensuring that the autonomy stack is UMAA compliant and can quickly be transitioned to physical hardware. This involves generating scalable digital twins for large multiagent systems, a composable high-fidelity simulation environment that can be tuned for a specific set of platforms, a digital twin-enabled communications optimization approach, resilient collaborative maritime autonomy leveraging multiagent learning-based control, and digital twin-enabled explainability and predictability of autonomous systems.