BAA 22-01 synopsis for SAM.docx

NAICS CODE: 541715

FEDERAL AGENCY NAME: Department of the Air Force, Air Force Materiel Command, AFRL - Rome Research Site, AFRL/Information Directorate, 26 Electronic Parkway, Rome, NY, 13441-4514

BAA ANNOUNCEMENT TYPE: Initial announcement

BROAD AGENCY ANNOUNCEMENT (BAA) TITLE: Fight Tonight

BAA NUMBER: FA8750-22-S-7001

PART I – OVERVIEW INFORMATION

This announcement is for an Open, 2 Step BAA, with a partial Closed-Staggered, 2 Step BAA for Technical Area 1 (TA1) and Technical Area 2 (TA2), which is open and effective until 30 Nov 2026. Only white papers will be accepted as initial submissions; formal proposals will be accepted by invitation only. While white papers will be considered if received prior to 4 PM Eastern Standard Time (EST) on 30 Nov 2026, the following submission dates are suggested to best align with projected funding:

Technical Area 1 (TA1) and Technical Area 2 (TA2) are structured as a Closed-Staggered, 2 Step BAA. Therefore, for TA1 and TA2 ONLY:

The Government anticipates awarding two awards for Technical Area 1 and two awards for Technical Area 2. Therefore, white papers for TA1 and TA2 are due on 10 JAN 2022 and if requested, proposals are due on 25 FEB 2022. Please note that white papers will still be accepted for TA1 and TA2 after the 10 JAN 2022 date, however, the likelihood of funding being available is substantially reduced.

For all other Technical Areas:

The Government anticipates adding technical areas via BAA amendment in the future. All potential Offerors are requested to wait to submit white papers until the BAA announcement is updated with specific technology requirements and dates.

Offerors should monitor the Contract Opportunities on the System for Award Management (SAM) website at https://sam.gov/ in the event this announcement is amended.

The Fight Tonight Program Team anticipates hosting an Industry Day to provide interested offerors an opportunity to learn more about Fight Tonight activities. Due to the ongoing COVID-19 pandemic, the Industry Day is anticipated to be offered both in-person and virtually. Details regarding time, place, registration, format, and attendee access will be published on SAM shortly following the announcement publication proper.

CONCISE SUMMARY OF TECHNOLOGY REQUIREMENT: The Air Force Research Laboratory, Rome, NY, is seeking innovative research that revolutionizes air operations planning by combining Artificial Intelligence (AI)-driven planning with interactive gaming to significantly reduce the Air Tasking Order planning cycle. Human-guided AI will generate myriad potential courses of action and the gaming environment will allow operators to explore, cull, and evaluate combat plans, gaining insight into the future battlespace.

The over-arching strategy of this five (5) year open BAA is to quickly and efficiently execute research and development to deliver practical solutions to urgent problems. These efforts are anticipated to entail rapidly integrating existing technology into field testable, functional prototypes of solutions to military problems in two (2) major Focus Areas:

Technical Area 1 (TA1): Interactive Plan Refinement Technical Area 2 (TA2): Plan Gaming and Outcome Analysis

This strategy provides AFRL Rome, NY a solicitation tool with the flexibility to solicit white papers and proposals for Contract or agreements to perform rapid prototyping of technical solutions to meet compelling Air Force needs.

BAA ESTIMATED FUNDING: Total funding for this BAA is approximately $99M. Individual awards will not normally exceed 54 months with dollar amounts normally ranging from $3M to $40M. There is also the potential to make awards up to any dollar value as long as the value does not exceed the available BAA ceiling amount.

Any anticipated funding listed reflects estimated program funding only. This estimate is not a promise of funding. Funding is uncertain and is subject to change. Changes in availability may occur as a result of the exercise of Government discretion.

ANTICIPATED INDIVIDUAL AWARDS: Multiple Awards are anticipated.

TYPE OF INSTRUMENTS THAT MAY BE AWARDED: Procurement contracts, grants, cooperative agreements or other transactions (OT) depending upon the nature of the work proposed. In the event that an Other Transaction for Prototype agreement is awarded as a result of this competitive BAA, and the prototype project is successfully completed, there is the potential for a prototype project to transition to award of a follow-on production contract or transaction. The Other Transaction for Prototype agreement itself will also contain a similar notice of a potential follow-on production contract or agreement.

AGENCY CONTACT INFORMATION: All white paper submissions and any questions of a technical nature shall be directed to the cognizant Technical Point of Contact (TPOC) as specified below (unless otherwise specified in the technical area):

BAA PROGRAM MANAGER:

Lt Col Sean Carlson

AFRL/RIS

525 Brooks Rd Rome, NY 13441-4505 Telephone: (315)-330-2936 Email: sean.carlson.1@us.af.mil

Questions of a contractual/business nature shall be directed to the cognizant contracting officer, as specified below (email requests are preferred):

Amber Buckley Telephone (315) 330-3605 Email: Amber.Buckley@us.af.mil

Emails must reference the solicitation (BAA) number and title of the acquisition.

Pre-Proposal Communication between Prospective Offerors and Government Representatives: Dialogue between prospective offerors and Government representatives is encouraged. Technical and contracting questions can be resolved in writing or through open discussions. Discussions with any of the points of contact shall not constitute a commitment by the Government to subsequently fund or award any proposed effort. Only Contracting Officers are legally authorized to commit the Government.

Offerors are cautioned that evaluation ratings may be lowered and/or proposal rejected if proposal preparation (Proposal format, content, etc.) and/or submittal instructions are not followed.

PART II – FULL TEXT ANNOUNCEMENT

BROAD AGENCY ANNOUNCEMENT (BAA) TITLE: Fight Tonight

BAA NUMBER: BAA FA8750-22-S-7001

CATALOG OF FEDERAL DOMESTIC ASSISTANCE (CFDA) Number: 12.800, 12.910

I. TECHNOLOGY REQUIREMENTS:

The Air Force Research Laboratory is soliciting white papers under this Broad Agency Announcement (BAA) for research, development, integration, test and evaluation of technologies/techniques that revolutionize air operations planning by leveraging artificial intelligence, with an interactive game engine for building, rehearsing and assessing combat plans. This capability will enable combat planners to quickly reason about plan options and plausible futures, increasing planners’ confidence that they have crafted the best plan possible within the constraints of a highly dynamic conflict against peer adversaries.

The U.S. military air combat planning process has evolved to require hundreds of experts working in concert to develop the plan needed to conduct military operations and fulfill Commander’s Intent. The future operating environment is expected to feature more mobile and fleeing targets that demand greater agility in planning to support execution. Hostilities against a peer adversary will likely demand a high sortie rate due to the speed of conflict and a potential numerical disadvantage. Current planning processes are generally serial and provide limited opportunity for sufficient analysis of options, requiring time-consuming plan adjustment and repair as conditions evolve during execution. While current processes generate efficient plans under controlled conditions, the pace and scale of future conflicts threaten to exceed current planning pace and required flexibility.

Air Operations Center (AOC) planning is an alchemy of interpreting strategic guidance, traditional optimization of known quantities (e.g., fuel burn rate, weapons capacity, logistics requirements) and unpredictability (maintenance issues, adversary actions). While automation excels at optimizing over well-defined variables, humans are better suited to understanding the strategic imperatives and reasoning under uncertainty. However, humans are limited in cognitive throughput and are constrained in their ability to reason over many potential futures.

The Fight Tonight program seeks to accelerate the planning process and to allow the exploration of potential courses of action for the AOC Strategy and Combat Plans divisions. The Strategy Division in the AOC converts guidance to objectives, performs mission analysis, course of action development, analysis and selection. These processes will be enhanced by tools that can generate and assess the feasibility of more plans than humans are capable of without the aid of technology. The Combat Plans Division develops detailed execution plans that meet the constraints of the operating environment, performing the detailed analysis and generation of artifacts for Air Tasking Order (ATO) preparation and re-tasking. The complexity of these activities and the need for close collaboration are the main drivers for the 36 hours currently used to complete the Master Air Attack Plan (MAAP) process in the AOC. This complexity reduces agility in competition with peer adversaries, motivating new concepts of operations. The Government envisions one such concept in Figure 1, with the Fight Tonight program providing the interface for continuous planning and refinement of strategy and operational assessment. While they are also critical functions, ATO production and execution management are not considered as part of this solicitation.

Artificial Intelligence alone cannot meet this challenge, human insight is required to guide and interpret automated planning and analysis in order to confidently issue courses of action. A Human and Artificial Intelligence driven collaborative planning system, combined with an interactive game engine, provides a leap-ahead capability for human planners to act at pace and scale. The interactive game engine would serve as the foundation for plan development, rehearsal and assessment under uncertain and dynamic conditions and is envisioned as the primary method for users to interact with the overall Fight Tonight system. The goal is to provide users with a consistent visual interface and environment to explore and select an allocation of forces, play, pause, and rewind the attack plan to understand critical decision points, manipulate the plan in real-time, assess potential plan changes, and explore various hypotheses for possible operating conditions.

The Fight Tonight program will develop tools that enable human users to work in collaboration with Artificial Intelligence to conduct air combat planning within 4 hours and incrementally conduct re-planning in minutes, while simultaneously exploring the trade-off between multiple options and their expected outcomes. Additionally, Fight Tonight will develop tools that optimize the role of the humans and machines – humans provide deep insight and creativity, while machines excel at reasoning over constraints and generating a detailed analysis of available options.

Figure 1. Fight Tonight will enable AOC planners to rapidly respond to an evolving battlespace and Combat Command (COCOM) strategy at pace and scale required for a peer adversary.

Developing plans in coordination with Artificial Intelligence at the pace and scale of future operations will require a common infrastructure with the following capabilities:

· Formalization of the representation and data needed to feed AI models and analytics.

· Flexible and persistent data interfaces serving Fight Tonight modeling and analytical capabilities timely information about the operating environment, as well as consumer applications with modeling and analytical output from Fight Tonight.

· Interactive user interfaces for users to provide guidance, real-time feedback, and quickly communicate details of complex operating environments in real time.

· Computationally scalable representations to manage state and update of large numbers of entities in an interactive, real-time environment.

· Capture and maintenance of gameplay for replay and analysis.

Commercial games have demonstrated the capability to meet these requirements but are generally developed as entertainment products with limited focus on planning courses of action. For example, commercial games like StarCraft II and Rome: Total War place the player in command of complex forces in real-time, providing multiple viewpoints for managing resources, maneuvers, and strategy. The Air Force Research Laboratory seeks to leverage the advances made in the commercial gaming industry, integrated with operational data and planning processes, to deliver a game-based planning engine for the Air Operations Center.

The envisioned game-based planning process will be non-serial, allowing users to develop an initial plan at coarse resolution and then iterate on finer levels of detail, with flexibility to return to lower fidelity planning as desired. It is envisioned that automated analysis will provide insight at all stages of planning, ranging from statistical analysis based on modeled uncertainty to execution of the plan using artificial intelligence to emulate friendly and adversary decision making. While Fight Tonight will build upon the well-established base of automated planning and scheduling technology, current capabilities are insufficient to meet the needs of the program vision and have limited integration with the data and framework required for continuous planning and outcome assessment.

The over-arching strategy of this five (5) year open BAA is to quickly and efficiently execute research and development to deliver practical solutions to urgent problems. These efforts are anticipated to entail rapidly integrating existing technology into field testable, functional prototypes that extends the state-of-the-art in two key areas:

· Interactive Plan Refinement. The ability to generate an executable plan via a user-guided, exploratory process of resource allocation and task prioritization. This service will aid the user to design allocations of forces to tasks, understand what options are available, and assess the impact of adjusting priorities. The proposed solution must be able to consider critical factors such as levels of acceptable risk, rules of engagement, friendly order of battle, enemy order of battle and the Joint Integrated Prioritized Target List (JIPTL).

· Plan Gaming and Outcome Analysis. The ability to interactively simulate execution of attack plans, assess variance of outcomes based on uncertainty, and generate empirical analysis of plausible futures. Planners should be able to explore and understand how a plan will unfold against a simulated adversary employing a variety of strategies and/or responses. The proposed solution should also support modification of plan elements, operating conditions, and capabilities for both friendly and adversary forces to determine how those changes will impact plan effectiveness and risk.

These technical challenges will be addressed as a two-phase program effort directed towards the development and evaluation of an integrated system that will be evaluated in a realistic military environment. The first phase of the program will focus on building a compelling technical demonstration of the Fight Tonight system, with the second phase intended to transition the developed capability for operational use in an AOC. The resulting planning system will need to address the inherent challenges of planning and analysis on fixed time periods, as well as integration with the human-guided automation to consider the cognitive needs of the human directing the planning process. In order to facilitate possible integration into larger enterprise platforms (e.g., Kessel Run), a robust software architecture will be needed to ensure Fight Tonight data and metadata is exposed and consumable by other applications and services (e.g., planners, gaming engines).

1. Program Structure Interactive plan refinement and plan gaming and outcome analysis comprise the two technical areas (TAs) of Fight Tonight development efforts. Figure 2 provides an overview of capabilities envisioned for each of the technical areas and how they are intended to interact, and Section 1.3 provides an overview of the objectives for each phase.

Figure 2. Notional functional components for the Fight Tonight architecture, and their association to Technical Areas requested in this BAA. Acronym Definition: Machine Learning (ML)

1.1. Technical Area Overviews

1.1.1. Technical Area 1: Interactive Plan Refinement

Interactive plan refinement will provide planners with tools to explore options for achieving mission goals through an iterative process. The system will initially produce coarse-grained, plausible plans based on user-provided guidelines on prioritization of multiple tasks and/or constraints. Automated analytics will enable planners to rapidly assess generated plan alternatives, and the system will work in parallel to continue refining candidate plans and increase the level of detail using available computational resources. This iterative process will require automated planning methods that emphasize speed over precision early in the process, then adapting based on continuous planner guidance to refine plan elements to a sufficient level of detail for simulated execution. This technical area must address the following challenges:

· Plan Synthesis – an automated process that generates plausible, approximate attack plans based on user-supplied objectives, resource allocations, and constraints. Automation will provide planners immediate feedback on plan feasibility in real time as they adjust allocations, informing further decisions.

· Plan Diversity Analysis – generates multiple plans based on varying prioritization of mission tasks and constraints and provides analytical assessment of the significant differences in performance and risk between the available options. This will free the user from having to inspect all options in order to understand the inherent trade-offs.

· Iterative Planning – an automated process for refining plans towards an increasing level of fidelity, allowing for early planner assessment and exploration of options before more computationally demanding plan elements are developed. This process may be interrupted in real-time as plan options are eliminated or additional guidance is provided by human planners.

· Plan Conditioning – brings plans to a sufficient level of detail for simulated execution within the TA2 gaming environment. This will help to bridge the analysis gap between plans and outcomes and provide additional confidence that a plausible plan is executable.

By addressing these challenges, Fight Tonight will develop capabilities to help human planners assess the effectiveness of plan options with respect to objectives at the increased pace and scale of future operations. There is no user interface anticipated for TA1 (apart from configuration and administrative function), with interactions between TA1 and TA2 components being defined through a Planning Interface that will be developed via participation in the Planning Interface Working Group (see Section 1.3 for additional details).

1.1.2. Technical Area 2: Plan Gaming and Outcome Analysis

The gaming platform will provide deep insight and experience with combat plans, allowing planners to analyze expected performance and update plans based on simulated outcomes. This is a key innovation that seeks to leverage commercial gaming technology to build intuitive, responsive interfaces that are designed to maximize human understanding of complex operating environments. Interface capabilities include faster-than-real-time plan execution, visual representation of spatial and temporal events of interest that have impact on plan outcomes, and methods for depicting where plan outcomes diverge or converge based on events. This technical area must address the following challenges:

· Gaming Interface - a human-centered, interactive set of visual interfaces for building, rehearsing, and assessing combat plans, driving and leveraging automation underlying the Fight Tonight system of systems. These interfaces will aid planners in contemplating the plausibility of force allocation and task prioritization options, the performance of the resulting plan against a simulated adversary, and the implications of hypothetical operating conditions.

· Plan/Analytics Data Store – a data storage solution with indexing across related plans, variations, assessment, and outcomes. Data representations and storage will need to account for the continuous refinement of plan options and the need for analytical techniques to access multiple plan options simultaneously.

· AI Behavior Models – automated players for friendly and adversary forces, enabling simulated plan execution and response to events. This will drive analysis of outcomes by executing plans over a range of conditions and uncertainty to support statistical analysis and visualization.

· Planning Execution Engine – a game environment for plan execution enabling analysis of expected plan outcomes under uncertainty and against varying adversary strategies. Adversary and friendly forces should be able to execute autonomously, driven by AI players of varying strategy and skill. The environment allows planners to explore potential futures, further refining the pace and timing of effects, constrain operating parameters, and identify opportunities and vulnerabilities that could emerge during conflict.

· Plan Analytics Engine – automated analytics to aggregate data on the various outcomes and present the planners with an overall assessment of the resulting outcome space. A visual depiction of these results will aid planners in focusing their efforts on the most likely and worst-case outcomes as part of the overall risk assessment process.

· Auto Exploration – automated exploration of adjustments to friendly and adversary resources, capabilities, and constraints to determine what impacts those changes will have on expected outcomes. Significant changes may require re-engaging with TA1 planning capabilities to generate new plans, with or without planner interaction.

This technical area is inherently human-driven, and development of interfaces and functional elements are driven by human decision processes. The decision space covered is both large and complex, introducing challenges for human cognitive bandwidth and allocation of analysis to a team of planners. The TA2 gaming interfaces will be the primary means by which planners interact with the system, providing a seamless progression from the exploration of plausible option, to understanding the potential outcomes based on uncertainty and adversarial strategy.

1.2. Evaluation Support

A Government-led team (composed of Government, Federally Funded Research and Development Centers, and potentially University Affiliated Research Center personnel) will design and conduct empirical performance evaluations throughout the life of the program. The Government intends to have the evaluation team provide a representative evaluation scenario and evaluation metrics to the technology development teams at kickoff. The evaluation team will continue to work with TA 1 and 2 performers during Phase 1 to provide an evaluation framework that will assist the development and assessment of component technologies. During Phase 2 and beyond, the evaluation team will work directly with the TA 1 and 2 performers to connect the integrated Fight Tonight capability to operational data sources. TA 1 and 2 performers will not integrate prototypes with Government systems during Phase 1 – instead the Government-led team will evaluate TA 1 and 2 capability demonstrations throughout Phase 1 and the proposed Fight Tonight system designs at the end of Phase 1.

The evaluation team will provide the following data inputs to TA1 and TA2 performers for development and demonstration in Phase 1:

· Basing: The physical locations where available air units and aircraft are located.

· Aircraft Models: The set of parameters defining the available resources used by the planner (e.g., fuel capacity and burn rates).

· Standard Configuration Loads (SCLs): Specific valid combinations of equipment and munitions for each aircraft type

· Air Units: An air unit maintains a group of aircraft of the same type at an air base. Multiple units may be located at the same base.

· Unit Contracts: Each air unit has a unit contract that defines how many sorties it will provide each day using its available aircraft. The unit contract also defines time windows during which the sorties can occur, and other constraints.

· Airspace Control Measures: The geographical coordinates of relevant air space regions like orbits, air refueling tracks, battlespace boundaries, etc.

· Targets and Target-Weapon Pairings: The prioritized list of targets (with location, description, etc.) and the munitions (type and quantity) recommended for use against them.

· Non-Strike Mission Requests: Additional requests for orbit based, ground alert, or other types of missions requiring aircraft support.

· Red Order of Battle: The estimated physical locations of adversary units and bases, as well as their capabilities against friendly forces. While some unit locations will be known, planning under uncertainty is expected to be a critical aspect of this program and the evaluation.

1.3. Program Phased Effort

To successfully demonstrate continuous planning with collaboration between human planners, automated planning, and interactive gaming platforms, AFRL expects all program performers to participate in refining the problem definition, identifying system requirements, and establishing interfaces, etc. At program start, the evaluation team will lead the effort of refining the functional architecture in coordination with the Government-led Strategic Planning User Group (comprised of operational users and subject matter experts) and Planning Interface Working Group (comprised of the Government-led evaluation team and representatives from each of the performers) that will be established at program kick-off and continue through the life of the program.

The Fight Tonight program will be conducted over a four-and-a-half-year period, in the following two phases:

1.3.1. Phase 1 (30 months)

The first phase of Fight Tonight will focus upon ensuring that the available data and technology being developed in TA1 and TA2 are comprehensive and realistic in meeting the needs and requirements of air campaign planners and relying upon data and knowledge that can reasonably be expected to be available to Fight Tonight applications. TA1 and TA2 development in this phase will be focused on ensuring that the program meets the needs of AOC planning staffs operating in a realistic environment. There will be strong emphasis on integrated capability demonstrations, with a focus on defining extensible and flexible interfaces and data standards. This includes identifying gaps between the needs of Strategy and Combat Plans Division staffs (gap identification will be conducted through the Strategic Planning User Group, with participation from all performers) and capabilities being developed in TA1 and TA2 (via the Planning Interface Working Group). There should also be a heavy emphasis on human factors system design to ensure that strategic and operational planning capabilities enable users to retain the insight and control provided by current manual processes. All efforts will focus upon framework design and architectural prototyping, testing the internal interface specifications, and the identification of interfaces between Fight Tonight internal and external components.

Figure 3. Fight Tonight Phase 1 iterative assessment process.

During this program phase, data artifacts and execution scenarios will be developed (by the Government-led evaluation team) for performers to evaluate the performance of their capabilities independently of the other Fight Tonight components. To facilitate this assessment, AFRL is planning a user-focused, iterative series of evaluation, depicted in Figure 3. At the culmination of each evaluation, all TA1 and TA2 capabilities will be assessed against the evolving test scenario and data set, with a goal of ensuring that the input/output requirements of the TA1 and TA2 capabilities are aligned. This process is intended to ensure that the TA1 and TA2 performers understand the needs of potential users, that their input/output assumptions are realistic, and that the scenarios being developed are sufficient to support the research and development and evaluation requirements for each TA1 and TA2 performer. This iterative refinement process also reflects the expectation that it is going to require substantial engineering effort in order to provide TA1 and TA2 performers with the data they need to develop and refine their technology. Note that it is expected that interfaces will converge through the iterative refinement process, informed by integration events, assessments, and the program-wide Planning Interface Working Group.

With the anticipated co-evolution of technology and concepts of operations at the Air Operations Center, the Government-led evaluation team will continue to identity and define key technical objectives, and their relative impact on program scope. The Government-led evaluation team will derive an analysis plan from these technical objectives in order to assess how well these objectives are being met. Based on this analysis, and in coordination with the Strategic Planning User Group, the Government-led evaluation team will develop a data collection plan that outlines what data will be collected in order to satisfy the analysis plan and measure to meet program objectives.

The Fight Tonight program will focus upon the integration and evaluation of capabilities in a functional prototype deployed in a realistic Air Force context. Components will be assessed for their ability to support planning staff to rapidly build, rehearse, and assess combat plans. Two major evaluation events with operational stakeholders are anticipated at roughly 12 and 24 months into the first phase, followed by a Phase 1 capstone demonstration in month 30. These evaluation events assess the potential for performers to be successful in Phase 2, based on metrics outlined in Section 1.6.1.

1.3.2. Phase 2 (24 months)

The final phase of the program will be driven towards a compelling demonstration of Fight Tonight’s ability to support air operations planning in a major Air Force evaluation context, to include planning for thousands of targets as part of a coalition force. An integrated prototype will be linked to the external systems necessary to support full-scale scenarios developed and conducted in close cooperation with representatives from operational and acquisition communities within the Air Force.

All proposals should address Phase 1 and Phase 2 development and clearly delineate effort in each phase, aligning to schedule and milestones provided in Table 1.

1.4. Technical Area Work Plans

1.4.1. Technical Area 1

The interactive plan refinement is intended to be iterative, with increasing fidelity and breadth of response as the analysis progresses. Initially, a planner seeks to determine if a plausible plan exists for a breadth of task prioritizations and allocations of forces to tasks provided by the planner through a specified TA2 interface. In parallel, the system continues to refine candidate plans to bring them closer to executable form and searching for additional plan options that maximize different objectives while meeting the specified constraints. The content and status of these plan options are continuously sent to the TA2 planner interface and provided to the user in visual form to aid in their assessment and further refinement.

For Phase 1 of the Fight Tonight program, human planner intent will be specified as a high-level set of objectives and constraints, a general apportionment of forces, and a prioritized list of potential targets, in a digital format. In this context, TA1 is devoted to creating the technology necessary to ingest data provided at run time and provide a service-based optimization engine that can generate feasible plan options and the associated analytics in real-time, as a component of a responsive and interactive user interface. These automated processes may be interrupted or altered by the planner at any time as they consider new options, eliminate options, and adjust or refine candidate plan options in the system. Therefore, it is critical that proposed planning technology is flexible and continuous, opposed to traditional methods focused on “solving” a single problem before providing the user with the results.

In general, key inputs for the services under TA1 include data on target priorities, associations of targets or target types to a task, multiple task prioritization options, adversary and friendly unit locations and status, airspace designations, valid aircraft standard configurations, and pairing of weapons with targets, including the associated probability of successful target effects. Specific details, including the size, frequency and means of delivery of this information will be determined as a part of the iterative concept refinement process planned for Phase 1, but the overall intention of the program is for TA1 to enable rapid TA2 human-guided planning and assessment to explore the possibility space.

AFRL anticipates making two TA1 awards and will likely continue with a single performer through the end of Phase 1, based on the performer’s ability to generate plausible coarse-grained plans from user input, iteratively refine plans into an executable form, support the need for interactive human-guided planning, and the integration of capability with the TA2 data interface. In order to enable a down-select of performers, the Government may request that proposals include separately priced options for Iteration 1 (months 13 through 24 of performance) and Iteration 2 (months 25 through 30 of performance). Proposals to this area may use representations that make sense from the perspective of their technology. However, these representations will ultimately need to be shared across the set of Fight Tonight functions, and the Planning Interface Working Group will be responsible for converging on a standard representation that can be used by all Fight Tonight performers. Thus, TA1 proposers should plan to work with TA2 performers via the Planning Interface Working Group to ensure that the standard developed supports their requirements. TA1 proposals should address the following capabilities when submitting a proposal:

1.4.1.1. Plan Synthesis

Given a prioritization of tasks and allocation of resources (e.g., aircraft, munitions, equipment), TA1 technology should respond with a plausible, course-grained plan for of delivering effects on targets based on available munitions, equipment, estimated flight duration and associated fuel requirements. This service is expected to be called in real-time to feed user interaction, prioritizing response time over accuracy in early interactions, with faster responses ideal for rapid development and assessment of plan options by the user. In order to meet interactive timelines, TA1 technology will likely utilize approximations of certain plan elements in order to reduce problem complexity for rapid response on plausibility queries, with iterative increases in fidelity in order to generate plan options and the correlated analytics. The ability to manage the demanding time constraints for interactive planning is a key challenge for rapid plan optimization.

1.4.1.2. Iterative Planning

Each plan option should be able to iteratively evolve from a plausible, course-grained plan to a fleshed-out plan of sufficient detail to be executed within the TA2 gaming environment. This process will be critical for meeting the pace and scale requirements of future operations. Proposals must address the strategy for enabling this capability, including identification of specific plan elements that are likely to require the most computation, plan elements that are most amenable to approximation and iterative refinement, and a methodology for user-guided changes of plans during the refinement process. It is also critical that the iterative refinement process takes place in parallel with other user activities, and on multiple plan options simultaneously, motivating approaches for managing limited computational resources.

1.4.1.3. Plan Diversity Analysis

Options for task prioritization and resource allocation form a complex decision space that will likely be too large for manual assessment by planners. Managing this complexity will be aided by data and analytics that compare the trade-offs of effectiveness and risk for each plan option, which will be presented visually in the TA2 user interface. As additional options are generated or existing options are refined, these analytics should be updated in real-time and presented to the planners to aid in the decision process. Proposals should address the need for analytical techniques, the resulting output, and methods for reflecting iterative changes in multiple plan options. Because the plan options are quickly evolving and may be fleeting, there is a need for rapid execution of analysis, emphasizing timely delivery of insight over high accuracy.

1.4.1.4. Plan Conditioning

The gaming of plan options is a key innovation of the Fight Tonight program, requiring that plan options eventually become executable plans that can be assessed based on simulated outcomes. To meet this requirement, plan options will need to be at a sufficient level of fidelity to be ingested by the Plan Execution Engine in TA2 through the Planning Interface and with data standards determined by the Planning Interface Working Group. The specifics of this level of detail are unknown at the time of this solicitation, but proposals should address methods to assess plan sufficiency and identify specific risks that may be incurred due to this requirement. Additionally, planners may propose changes to plans within the gaming environment that violate key dependencies or constraints, requiring re-engagement with the iterative planning process for that plan option. Methods for assessing the need to re-engage in the plan development process are encouraged, but the nature of this interaction will not be clear during early technology development.

1.4.2. Technical Area 2

The goal of TA2 is to develop and demonstrate the Fight Tonight interactive system, including the software infrastructure, user interfaces, and planning workflow. This includes support of evaluations and demonstrations through the life of the program and facilitating transition. The relatively short timeline to complete these complex tasks necessitates a seamless transition between modes of operation and motivates the TA2 performer being the integrator across these capabilities.

A key innovation under Fight Tonight is the use of a gaming engine to provide planners with an interactive capability to explore and refine combat plans. Commercial games have demonstrated a capability for visual and analytical abstraction of complex operating environments combined with real-time interaction to assess and control forces under the users’ command. For Phase 1 of the Fight Tonight program, the use of gaming as a tool for attack plan development will be demonstrated, including interfaces to the TA1 services, user interfaces and ingestions of provided data artifacts by the evaluation team. Specific details, such as the required timelines for plan assessment and data formats, will be determined as a part of the iterative concept refinement process planned for Phase 1, but the overall intention of the program is for TA2 to enable development, analysis and refinement of a MAAP within 4 hours by the end of the program. TA2 is intended to develop an interactive user interface with a consistent set of visual interface elements, creating a common user experience for the entire attack plan planning process.

The commercial gaming industry has evolved a mature set of tools for efficient design, implementation, and rendering of visual interfaces, so proposers should leverage these capabilities when possible, or provide justification on why existing tools are not appropriate for this task. The game interface will evolve based on feedback from the user community via the Strategic Planning User Group, stakeholder participation in program integration and assessment events, the capabilities of the TA1 optimization service, and data standards determined by the Planning Interface Working Group. This evolution necessitates a flexible and modular architecture for integration of the game engine components, data storage and retrieval, and user input into planning processes.

1.4.2.1. Game-Based Planner Interfaces

The development of a set of interactive graphical user interfaces will empower planners to quickly build, rehearse, and assess plan options. To support this vision, the program asserts that a set of visually consistent interfaces with shared user interface elements across all phases of the planning process will maximize the ability for planners to rapidly develop and refine combat plans. To realize this capability, the Government envisions three key interfaces that embody functionality with somewhat distinct, but inherently overlapping, scope and purpose.

Interactive Force Allocation The program envisions the first step of the planning process could be allocation of resources to meet specified objectives and constraints within the operating environment. This process is envisioned as an interactive “drag and drop” interface, with real-time system responses from the TA1 systems to assess the current strategy. The goal of this step is to understand if a plausible plan exists based on task priorities, constraints, dependencies, and available resources, and what other options may provide a suitable or preferred alternative. Plan options will be assessed by analytics provided in TA1, which will be presented to the user in an integrated interface during the interactive planning process.

Planning Gaming Engine Once a plausible plan option has been sufficiently refined into an executable plan, planners can execute the plan in a game environment to gain insight into the space of plausible outcomes. The program envisions options for the planners to play their plan against varying adversary AI opponents, play against another human planner or adversary subject matter expert, or direct a friendly AI and adversary AI to play against each other in order to sample a larger space of outcomes. Capturing the diversity (or similarities) of outcomes and presenting a concise depiction in visual form will be critical to support rapid decision making and further refinement.

Hypothesis Generation During or after plan execution in the game environment, planners may pose changes to the plan or the operating environment, with the system providing feedback on how those changes affect the currently viewed plan and associate completion of mission objectives. The changes could include added or reduced resources available to friendly or adversary forces, changes in airspace restrictions, additional basing options, or changes in friendly or adversary behavior to be executed by the AI players. The Fight Tonight program seeks approaches that will enable users to explore these changes as branches of the baseline plan, with options for committing changes and having them be part of the revised baseline.

1.4.2.2. General Game-Based System Capabilities

To realize the Fight Tonight vision, the program assumes there are a set of underlying capabilities that will be needed to drive the interaction between human planners and Artificial Intelligence. Note that these are intended to be capability descriptions and may not reflect the proposed functional architecture.

Plan Execution Engine Given a plan generated through interaction with the optimization service, Fight Tonight envisions providing planners with the capability to play through the execution of the plan in an interactive gaming environment. This will require the generation of the scene elements, mapping of plan actions to the respective units, and instantiation of AI behaviors (as directed by the user). The data requirements for this capability are expected to be a critical discussion between TA1 and TA2 teams as part of the Planning Interface Working Group.

AI Behavior Models To enable automated play of friendly or adversary forces, AI models could be made available for general use or trained against plan instances, as time permits in the planning process. It is envisioned that these models could have some combination of “micro” AI for controlling individual units, and a “macro” AI for directing behavior of multiple units. Additionally, there would be value in having AI models of varying strategy, providing planners with an analysis of plausible plan outcomes against an uncertain adversary under varying conditions.

Auto Variant Exploration While human plan manipulations will be invaluable to explore the possibility space of plans and operating conditions, the program envisions an automated capability to explore variants of plans and operating conditions to identify potential opportunities or vulnerabilities. This will require a capability to search the space of plan adjustments (still adhering to critical constraints and dependencies) and assessing the changes in outcomes. The results of this analysis would ultimately rely on the human to determine the likelihood of the hypothetical scenario, the opportunity versus risk it entails, and if it warrants further exploration.

Plan Analytics Engine Given the elements of uncertainty for a plan, adversary response, or operating environment, simulated execution of multiple plan options provides insight into the space of possibility and the most likely outcomes. The program envisions the use of analytics to aggregate data on the various outcomes and present the planners with an overall assessment of the resulting outcome space. A visual depiction of these results will aid planners in focusing their efforts on the most likely and worst-case outcomes as part of the overall risk assessment process.

Plan & Analytics Data Store Facilitating game play of plans, managing manipulations and analytics, and presenting aggregate information to users requires a data storage solution with indexing across related plans, variations, assessment, and outcomes. Data representations and storage will need to account for the continuous refinement of plan options and the need for analytical techniques to access multiple plan options simultaneously. Additionally, data will need to be accessible by external services to continue the AOC planning process.

Software Architecture and Integration In order to reduce costs, avoid duplication of effort, and focus program resources on other high-risk development, TA2 proposals should focus upon assembling an appropriate set of existing commercially available, open source and government-owned software capabilities, leveraging existing message sets and standard interfaces to the maximum extent possible. Proposals to develop new capabilities should address the technology gaps being addressed to justify the need for custom solutions. Proposers to this task should clearly describe the software development process that they intend to use. Processes and data interfaces should be flexible enough to support the program need for integration of independently developed functionality, with enough rigor to produce reliable, tested software with adequate documentation.

As the systems integrator, the TA2 performers will be responsible for the integration of TA1 capabilities into the Fight Tonight prototype system that will be demonstrated during Phase 1 (and in the notional Phase 2) of the program. They will support technology evaluations by providing iterative versions of their software framework to TA1 researchers so that the Government can assess the performance of the TA1 services, as well as the framework’s ability to support that performance. It is critical that the TA2 performers create a common runtime environment that can support the flexible integration of software capabilities, including core infrastructure services such as communications and access to shared models and data stores.

As the integration lead for Fight Tonight, TA2 performers will need to be able to access secure facilities and design software solutions that can be used in classified environments, coordinating activities across multiple security levels. Consequently, proposers to TA2 must demonstrate the ability to access and store data at the TOP SECRET/SCI level. Finally, in order to facilitate transition, the TA2 performers should demonstrate knowledge of current and in development Air Force planning systems and demonstrate how their prototype middleware will enable integration into those environments.

AFRL anticipates making two TA2 awards and will likely continue with a single performer through the end of Phase 1 based on assessments of usability, responsiveness, scalability, and potential for operational transition in Phase 2. This decision will be based upon several factors, including assessments of usability and scalability, the overall system design, a proposed integration and risk mitigation plan for Phase 2, the performer’s ability to assist with technology transition, and the performance of a prototype developed during Phase 1.