Making “SENSE” in the Maritime Domain

With today’s geo-political conflicts developing on a worldwide scale necessitating the need for timely and accurate information by decision makers, the requirement for maritime Intelligence, Surveillance and Reconnaissance (ISR) has never been greater. Maritime ISR by itself is difficult due to the vast distances, dispersed activities and limited high-bandwidth communications involved.

As the continuum of conflict advances into grey zone or active hostilities, maritime operations become more difficult.

Limited access to denied areas, constrained reliable communications availability, and the need for quickly detecting developing patterns of behavior all place significant demands on maritime platform operations. Fielding of large numbers of naval surface vessels and submarines able to survive in this environment is costly and schedule prohibitive. The need for lower cost unmanned platforms to fill the gap has been well documented, however a rigorous analysis of optimal numbers, types, and Concepts of Operation (CONOPS) remains an open area of research.

There are some physical attributes that these unmanned maritime platforms share including the need for Persistence to operate in an area for long periods of time, Endurance to transit long distances and Reliability to continue functioning with some loss of component failure. There are also operational attributes these platforms share including autonomous sensing, processing and communications required to fulfill mission requirements. BAE Systems has been exploring this mission area from a platform and mission systems perspective with a particular focus on “sense making” from gathered information. The key objective is maximizing the amount of information produced with the minimum cost of force structure while also providing a degree of resiliency to loss of operational units through attrition or hostile action. In this article we draw comparisons to a neural network analogy for situation-dependent force deployment.

ESSENTIAL INFORMATION

Within the information processing / intelligence domain, the term “Essential Elements of Information” (EEI) is typically used to describe information needed to satisfy the operational commander’s priorities of intelligence requirements. Gleaning these EEIs benefits from persistence to gather the information as well as the sensing modalities required to obtain that information. Access to the information in contested areas, at great distances, can be overcome by the use of intrusive platforms (i.e. UUVs). Due to the distances involved and the limited number of platforms available, it is our belief that a heterogeneous mix of multi-modal sensors is most effective for covering the acoustic, RF and optical environments. EEI’s change over time and the UUV platform requires both short-term mission level sensor planning and autonomy as well as long term platform sensor and geographic (spatial) placement planning. The ability to make sense of what is being observed and communicating that to a Mission Operation Center (MOC) drives the requirements for processing and communications subsystems.

Borrowing from UAV operations, if high-bandwidth communications channels are available, processing can take place remotely (at the MOC for instance), however in constrained communications environments, processing has to take place forward at the edge. UUVs inherently have very limited high-bandwidth communications opportunities, thus energy efficient onboard processing is extremely important. This processing needs to address the fundamental questions of “Who? Where? And when?” implying the need for precise geo-positioning, timing and some means of Automated Target Recognition (ATR) acting on the sensor products.

A LAYERED NETWORK OF UUVs

To cover an area of interest we need multiple UUV platforms with sufficiently overlapping sensor fields of regard to achieve the persistence necessary to observe activities that may take place over long (days or more) periods of time. We can think of these forward-based sensing elements as nodes in a graph or input layer of a neural network. These sensing elements have tasking that may change over time (who, where, when) with each sensor modality having a weight applied to it. These weights could be scheduling of particular sensor modes or sensor prioritization schemes for instance. How these weights change by mission tasking could be assigned back at the Mission Operations Center (MOC) manually, or autonomously through the use of a middle layer of UUV platforms. This middle layer becomes important as its platforms are physically distant from the forward based sensor nodes, but have the time advantage to maneuver to add additional observations and data to the network.

As an example, if the forward sensor UUVs (nodes) are tasked to monitor the channels into a port facility, upon detection of a potential vessel of interest they can Tip and Cue the middle layer UUV(s) to intercept and provide additional information on who, where and when. This more fully refined information is then sent back to the output layer (Mission Operations Center) to answer why this behavior is occurring and what does it imply?

In summary, we have a layered network of UUVs working together to meet wide area coverage rates and long temporal duration collection requirements. EEI’s are satisfied through use of forward based sensor collection platforms and fine grain analysis of suspect platforms through middle layer of UUVs which collectively provide situational awareness to an operational center (output layer node). Different layers (UUVs) have different operational properties. Forward UUVs may only need to provide sufficient information to cue intercepting platforms (middle layer of UUVs), and can be configured with payloads accordingly. The middle layer needs the ability to perform some level of fine grain sensing and higher level of information processing (sense making) and communications. This offers the possibility of a High-Low mix of platforms where forward based sensor platforms are relatively inexpensive and middle layer platforms may have more expensive processing and communications capabilities.

ACTIVITY BASED INTELLIGENCE

As data is collected, Pattern of Life (PoL) analysis is performed over time at the MOC, abstracting detections and single-vehicle activities into higher-echelon threat behaviors, posture, and objectives. Information needs identified by PoL algorithms back propagate value into the weights of the multi-layer sensing network, geared toward gathering mission EEIs to refine threat estimates. For example, drug manufacturing, weapons smuggling, troop movements, and oil smuggling all have different precursor activities (Indications and Warnings or I&W) associated with them. A self-annealing autonomous UUV network and PoL engine could potentially self-cue or adjust weights and activity functions to determine what is taking place and how best to adjust the network to confirm the behavior, similar to neural network feed forward and back propagation techniques. Thus, the middle layer of UUVs may place emphasis (weight) on processing and communications over multi-modal sensing of the forward based sensing UUVs. The use of artificial intelligence / machine learning (AI/ML) to discover patterns of behavior through widely dispersed (temporally and spatially) activities is essential to future success. UUVs that can be reconfigured during a mission are also necessary to enable this dynamic unmanned systems self-annealing networked vision.

This framework of Activity Based Intelligence has been proven in the past. Ideally the first two layers of the network would operate autonomously and discover illicit behaviors on their own. A fully autonomous UUV neural network could potentially self-cue or adjust weights and activity functions to determine what is taking place and how best to adjust to confirm the behavior reducing the workload on analysts and operational planners at the MOC. As bad actors adjust or camouflage their behaviors, AI/ML techniques may provide a higher likelihood of detecting associations in a timely manner over traditional human analytics-based approach. Camouflaged behaviors also drive the need for multi-modal sensing approaches and clandestine approach/ proximity to the behavior itself.

ENABLING TRUE UUV AUTONOMY

The end state is to build a self-annealing network of UUV platforms that autonomously adapt their behavior to satisfy a mission role with minimal human intervention. The issues of energy regeneration and available communications means have not been addressed in this article. These requirements are being addressed through current navy programs. However, the need for forward based platforms to survive (maintain situational awareness and ability to react to perceived threats) and operate reliably through some means of energy regeneration needs to be addressed as part of the overall operational concept. Communications may be conducted asynchronously or within a window of availability, however all nodes in the system need to have the most recent copy of “weighting” or tasking in order for the current operations to be effective.

BAE Systems is currently conducting research in Maritime System of Systems architectures through a physics-based modeling and simulation environment and Model Based Systems Engineering approach. In addition to being a platform provider we are also a mission payload provider with ongoing research in wideband apertures, acoustic, optical and RF sensing, UMAA compliant autonomy systems, tracking and fusion, and AI/ML algorithm development for maritime applications. Lastly, BAE Systems has proven pattern of life capabilities for a wide range of threats and data modalities. Collectively, with key enabling partnerships within government and industry, we are ready to develop and deploy layered sensing networks to achieve persistent, reliable, and intelligent situational awareness in contested environments. 

This story was originally featured in the digital issue of ON&T April 2021. Click here to read more.