SAILS (Situational Awareness by Intelligent Learning Systems)

Problem Statement

U.S. Navy assets generate vast amounts of multi-source tactical data from various platforms, including space-based, shipboard, and airborne assets, as well as unstructured data (intelligence reports, watch logs, etc.) produced by sailors. Currently, Maritime Operations Centers (MOCs) must manage and analyze large volumes of multi-source data generated across the fleet to make critical resource allocation decisions for geographically dispersed fleet and national assets.

Desired Functional Attributes

The Navy seeks commercial AI/ML applications that accelerate the convergence of MOC-destined data inputs (e.g. intelligence reports, satellite-derived data, and existing common operational picture tools, etc.) to improve situational awareness for operators, and optimize existing decision support tools by offering track confidence scoring and real-time recommendations to assist commanders in allocating geographically dispersed resources (e.g. satellites, aircraft, vessels, etc).

Selected participants will work alongside government engineers and end-users while partnered with ashore operators. Companies are welcome to apply individually or in partnership. Vendors are required to include use cases of their solutions in decision support settings as part of their submission.

The following outlines the capabilities of desired applications. Submissions should provide examples of actual project work indicating the ability to develop and deploy such applications: 

1. Watchfloor Workflow Automation: 
   1. Connect to third-party software and data platforms through APIs to deliver data-driven and workflow automation models that are specifically developed for MOC use cases (e.g. validation rules to standardize data entry, automated routing of workflows to reduce manual mission approvals, dynamic updates for mission critical assets/resources etc)
   2. Vendors should/must offer past instances where their solutions implemented standardized API protocols (e.g. RESTful APIs with OAuth 2.0 authentication and HTTPS encryption) and data exchange schemas (e.g. JSON, XML, etc) 
2. Track Confidence Determination:
   1. Develop and deploy machine learning (ML) models to generate track confidence scores taking into account MOC-level user interests in platform/system-level health, historical accuracy of platform/system-level reporting, and ever-evolving operational contexts
   2. Propose methods to generate confidence scores and intervals for predictions that incorporate dynamic inputs (e.g. asset/platform health metrics, user inputs, etc) into the scoring architecture
   3. Deliver threshold-based alerts to end-users (e.g. when confidence scores fall below user-defined thresholds)
3. Sensor and Resource Optimization:
   1. Generate sensor and resource allocation recommendations - accounting for communication bandwidth conditions, geographic constraints, sensor reliability, past model performance, watchstander availability, etc. to inform MOC Commanders of asset/resource availability and readiness
   2. Vendors should describe how models can deliver a range of outputs based on time constraints, and allow end-users to dictate the degree of feasibility or optimality, depending on the time available to process the solution
4. Trust with the Warfighter; Modularity for the Force
   1. For the above applications, provide natural language-based model tuning that allows MOC end-users to interactively adjust objective functions, factors, and constraints (e.g. asset/resource downtime, readiness, and availability, etc), while ensuring that the model’s decision-making process is maximally interpretable and/or explainable
      1. Vendors should/must cite their demonstrated ability to deliver interfaces that expose model parameters to end-users to allow the modification of mission-specific factors (e.g. downtime thresholds, readiness criteria etc);
      2. Vendor interfaces should also enable end-users to adjust constraints (e.g., bandwidth limits, sensor/asset reliability weights), and explain how adjusting constraints impacts outputs (e.g. resource allocation)
      3. Vendors must supplement specified methods with training modules to establish and maintain trust with end-users
   2. Offer modular, open-architecture AI applications capable of integrating into existing and future DoD systems and interfaces. Application designed with modular components such that core logic, data ingestion and output, and user interface are decoupled to support users and workflows across DoD
   3. All Proposals must align with DIU's Responsible AI Guidelines and other DoD AI frameworks. Companies are encouraged to read the guidelines and develop a plan for implementation, which may be presented as part of the submission.

Desired Technical Attributes

• Enable role-based access control (RBAC) and cross-domain data sharing
• Support containerized, hardware-agnostic deployment
• Ability to operate to the maximum extent possible in Denied, Disrupted, Intermittent and Limited Impact (DDIL) bandwidth areas
• Solutions should support deployment on government or contractor-provided infrastructure and allow for operations across different classification levels (IL5/IL6/IL7) while preventing unauthorized access to data
• Solutions must comply with NIST 800-171.
• Provide documented capabilities, limitations, and explainability for developed models

Companies must have experience deploying on classified networks and ability to perform work in these environments and cleared personnel who can work up to TS/SCI (teaming arrangements will satisfy this requirement). It is encouraged that submissions highlight experience with controlled environments, such as deployments in GovCloud or any classified data/environments (https://public.cyber.mil/dccs/). Preference will be given to companies whose solutions have been deployed on IL6 or higher or have an equivalent Authority to Operate (ATO).

Companies are expected to demonstrate their solution in an unclassified environment as part of Phase 2 of the Commercial Solutions Opening.

The Government may seek to team companies with complementary solutions under the umbrella of this AOI; or in coupling of other AOIs’ issued within the DIU. Selected vendors may be required to engage in partnerships with other vendors to achieve Government objectives.