The Department of Defense, specifically the United States Air Force, is seeking innovative solutions through the Small Business Innovation Research (SBIR) program for the development of an AI-driven Team Manager and Translator (ATMAT) software tool. This initiative aims to enhance teamwork and coordination within Joint All Domain Command and Control (JADC2) operations, particularly for fluid teams operating in distributed Multiteam Systems (MTS) under denied and degraded communication conditions. The proposed software will leverage advanced technologies such as interactive machine learning, knowledge graphs, and large language models to create dynamic knowledge databases and predictive models that facilitate real-time situational awareness and decision-making. The opportunity is currently in the pre-release phase, with an open date set for May 28, 2025, and a closing date for applications on June 25, 2025. Interested parties can find more information and submit proposals through the official SBIR website at https://www.dodsbirsttr.mil/topics-app/.

## Description

Future JADO and CJADC2 operations will necessitate the formation of fluid teams, composed of operators from diverse backgrounds and expertise, who must rapidly integrate into temporary, distributed MTS. These fluid teams, characterized by limited shared experience and potentially restricted communication, often face shortcomings compared to stable teams due to challenges in rapidly developing teamwork structures like shared mental models, transactive memory, and calibrated trust. Additionally, the complexities of MTS, with varying levels of functional interdependence and integration mechanisms, further complicate coordination efforts. Finally, the need to coordinate and collaborate in denied and degraded communication environments necessitates coordination devices that allow rapid synchronization of knowledge within and across echelons and MTS and for real-time and accurate prediction of MTS state enabled through modeling of MTS dynamics and conditions-based authorities.&nbsp;The resolution of these issues will require a detailed model of MTS state and dynamics that is scalable, adaptable, and compatible with established and emerging AI frameworks. In particular, this problem space calls for a comprehensive solution that can create dynamic knowledge databases and predictive models that leverage existing documentation and information to contextualize and understand ongoing team and system dynamics. Software solutions are needed that can ingest, parse, and process disparate sources of potentially unstructured information into standardized representations that can model entities, properties, and relationships and use these as a priori models to ground and contextualize team communications, functioning, and dynamics.&nbsp; Importantly, the system will need to be scalable such that higher-order entities (e.g., teams) that comprise combinations of lower-order entities (e.g., teammates) can be modeled with properties that are more than the aggregate of the individual parts. This concept should be scalable to individuals, teams, MTS systems, operational systems, etc., with capabilities to model unique properties and dynamics at each scale of observation. Additionally, individual elements should also retain identity and properties so that changes in the higher order meta-stable properties and relationships can be traced to root causes.&nbsp;The system should be modular, with accessible APIs and standard libraries to allow tailored use of interchangeable LLMs, retrieval augmented generation, scalable databases, along with intuitive GUI-implemented manual and human-in-the-loop interactive machine learning capabilities for updating and augmenting all models, documents, and other relevant contextual information. This GUI interface should be usable by those with little to no computer programming experience. Finally, in addition to grounding and predicting MTS states, the system should be able to leverage generative AI operating on the knowledge graphs and models of MTS communications and dynamics to create forecasts of team performance and possible interventions, as well as synthetic data for data augmentation capabilities and null hypotheses distributions to allow statistical inference about observed MTS dynamics.&nbsp;The proposed software tool aims to augment teamwork and optimize coordination within these fluid MTS environments by providing team members with accessible information about each other's relevant experiences, knowledge, skills, abilities, and platform capabilities, integrated into a comprehensive knowledge graph framework. The system will also represent the dynamics of knowledge across temporal and spatial dimensions within the modeling framework, enabling AI reasoning about changes in the operational environment and team states over time to predict teaming outcomes and assist in ensuring distributed situational awareness by generating real-time alerts, predictions, and pushing critical information to the appropriate level at the appropriate time. This will ultimately enhance decision-making and responsiveness through AI managed distributed situational awareness.&nbsp;

## Objective

Develop a software tool/platform to enhance teamwork and coordination in Joint All Domain Command and Control (JADC2) operations, specifically addressing challenges associated with fluid teams operating within distributed Multiteam Systems (MTS) in denied and degraded communications environments. The platform will leverage interactive machine learning, knowledge graphs, and large language models to generate and update operationally relevant representations of individual and team states and dynamics and to integrate and reason over disparate forms of data, enabling advanced features like real-time alerts, predictions, and intelligent information dissemination.

## Phase I

The primary objectives of Phase I are to develop a conceptual framework for a scalable and adaptable knowledge management and representation framework for distributed multi-team systems. This system will leverage LLMs, RAG, knowledge graphs, interactive machine learning, and graphical user interfaces. The framework's scalability and transferability will be conceptually demonstrated across different domains through incorporating pipelines and procedures for easily augmenting and annotating databases, implementing retrieval augmented generation techniques, tailored use of LLMs, and providing intuitive interfaces for interactive machine learning. These interfaces will facilitate the generation, verification, maintenance, and updating of the underlying knowledge representations and generative AI framework and should be accessible to both those with little to no programming ability (e.g., through graphical user interfaces) and to developers (e.g., through APIs, accessible databases, and libraries). Furthermore, an analysis of fluid team dynamics and MTS coordination and communication challenges within JADC2 operational contexts will be conducted, with a specific emphasis on dynamic targeting or kill-web operations to inform targeted model development.&nbsp;This phase aims to develop a conceptual framework for the software tool, outlining its key functionalities and integration with existing or novel experimental MTS systems and testbeds. The expected outcomes of Phase I include an initial conceptual framework and a demonstration of the technical feasibility of the proposed software tool. Additionally, this phase will describe initial AF-relevant use cases, including the application of the framework to previously collected data and its integration into existing or novel MTS experimental testbeds capable of simulating denied or degraded communications environments, ideally incorporating dynamic tactical, operational, and strategic elements. Finally, Phase I will identify key individual and teamwork performance indicators and metrics for subsequent phases.

## Phase II

Phase II objectives include implementing the conceptual framework from Phase I and conducting a full cognitive task analysis (CTA) of team interdependencies within the identified MTS use cases. A prototype will be designed and implemented to demonstrate core capabilities, such as incorporating CTA results into the modeling framework, constructing knowledge graphs from diverse data sources (including USAF and DoD doctrine, team communications, task performance data, etc.), and showcasing real-time updates and reasoning in a use case identified in Phase 1. The framework's scalability and transferability will be demonstrated by analyzing multiple data sources across at least two domains, and a suite of basic alert and prediction models will be created based on knowledge graph analysis and models of team effectiveness. Prototype effectiveness will be routinely evaluated and refined by testing the ability of subject matter experts with little to no programming experience to rapidly build out annotated data sets and models of MTS dynamics, through evaluation of generated models for evaluating, predicting, and improving MTS functioning and dynamics in Air Force-relevant scenarios, and through user feedback testing these models.&nbsp;The expected outcomes of Phase II are a well-defined, user-friendly prototype demonstrating streamlined data annotation and model building, significant improvements in fluid team performance and MTS coordination, validation of the software's effectiveness in enhancing situational awareness and decision-making, and readiness for transition into operational environments.

## Phase III: Dual Use Applications

Phase III objectives encompass the continued development and maturation of the framework established in Phase II. Leveraging insights gained in the previous phase, the interface, capabilities, and components will be refined to facilitate the transition of the technology to both military customers and the open market. The software tool will be demonstrated in high-fidelity training or operational use cases within JADC2 or similar environments. Rigorous testing and evaluation will be conducted in realistic operational settings, involving representative user groups, followed by further testing in real-world scenarios. The framework will be expanded and developed to effectively handle real-time information flows and evolving contexts. Additionally, plans and mechanisms will be provided for seamless integration with existing JADC2 or other operationally deployed systems and platforms. Comprehensive training and support will be offered to end-users to ensure successful adoption and utilization. Expected TRL level at Phase III entry is TRL 7. Potential transition partners include ACC/A307 and A326 and this will be enabled by leveraging existing complementary programs with likely paths for transition to these customers.

AI-driven Team Manager and Translator (ATMAT)

SBIR Opportunity Analysis

SBIR Documents

Related SBIR/STTR Opportunities

Opportunity Snapshot

Key Dates