HR001119S0051.pdf

Broad Agency Announcement

Air Combat Evolution (ACE)

STRATEGIC TECHNOLOGY OFFICE

HR001119S0051

June 5, 2019

TABLE OF CONTENTS

PART I: OVERVIEW INFORMATION

PART II: FULL TEXT OF ANNOUNCEMENT

I. Funding Opportunity Description

A. Program Goal B. Program Overview C. Schedule and Deliverables D. Program Metrics

II. Award Information

A. General Award Information B. Fundamental Research

III. Eligibility Information

A. Eligible Applicants B. Organizational Conflicts of Interest C. Cost Sharing/Matching D. Other Eligibility Criteria

IV. Application and Submission Information

A. Address to Request Application Package B. Content and Form of Application Submission

V. Application Review Information

A. Evaluation Criteria B. Review of Proposals

VI. Award Administration Information

A. Selection Notices and Notifications B. Administrative and National Policy Requirements C. Reporting D. Electronic Systems

VII. Agency Contacts VIII. Other Information

IX. APPENDIX 1: PROPOSAL SLIDE SUMMARY

X. APPENDIX 2: VOLUME 1 COVER SHEET TEMPLATE

XI. APPENDIX 3: VOLUME 2 COVER SHEET, CHECKLIST AND SAMPLE TEMPLATES

XII. APPENDIX 4: SECURITY CLASSIFICATION GUIDE REQUEST FORM

XIII. APPENDIX 5: ASSOCIATE CONTRACTOR AGREEMENT (ACA)

PART I: OVERVIEW INFORMATION

Federal Agency Name – Defense Advanced Research Projects Agency (DARPA), Strategic Technology Office (STO) Funding Opportunity Title – Air Combat Evolution (ACE) Announcement Type – Initial Announcement Funding Opportunity Number – HR001119S0051 Catalog of Federal Domestic Assistance Numbers (CFDA) – Not applicable Dates o Proposers’ Day – May 17, 2019 o Posting Date – June 5, 2019 o Questions Due Date and Time – June 26, 2019, 5:00 PM (EDT) o Notification due to DARPA PSR if proposing any classified information – June

28, 2019, 5:00 PM (EDT)

o Proposal Due Date and Time – July 22, 2019, 5:00 PM (EDT)

Total amount anticipated to be awarded – Up to $63.6 million (Phases 1, 2 and 3) Types of instruments that may be awarded -- Procurement contract or other transaction.

Anticipated individual awards – Multiple awards are anticipated.

Agency contact

The BAA Coordinator for this effort can be reached at:

electronic mail: HR001119S0051@darpa.mil.

Or USPS mail:

DARPA/STO

ATTN: HR001119S0051

675 North Randolph Street Arlington, VA 22203-2114 mailto:HR001119S0051@darpa.mil

PART II: FULL TEXT OF ANNOUNCEMENT

I. Funding Opportunity Description

This publication constitutes a Broad Agency Announcement (BAA) as contemplated in Federal Acquisition Regulation (FAR) 6.102(d)(2) and 35.016 and 2 CFR § 200.203. Any resultant award negotiations will follow all pertinent law and regulation, and any negotiations and/or awards for procurement contracts will use procedures under FAR 15.4, Contract Pricing, as specified in the

BAA.

The Defense Advanced Research Projects Agency (DARPA) is soliciting innovative proposals for the Air Combat Evolution (ACE) program. The ACE program will increase warfighter trust in combat autonomy by automating aerial within-visual-range (WVR) maneuvering, colloquially known as a dogfight, using progressively realistic platforms (following demonstrations in modeling and simulation, the program will graduate to small-scale unmanned aerial vehicles, and culminate on operationally representative aircraft). A dogfight represents the progression in autonomy from current physics-based automation generally trusted by operators to more complex human-machine collaboration necessary to realize the promise of future manned/unmanned teaming.

A. Program Goal

DARPA is developing a new warfighting concept called Mosaic Warfare, which is an approach to combined arms maneuver, wherein capabilities traditionally provided to close a monolithic kill chain dependent upon a single, eminently capable platform are instead provided by a heterogeneous set of manned and unmanned systems. When properly orchestrated, these heterogeneous kill chains deliver a range of necessary effects.

In the Mosaic Warfare vision, humans are expected to fight in close collaboration with autonomous weapon systems in complex environments (such as those described by coupled, nonlinear, heterogeneous, and adaptable agents) with tactics informed by artificial intelligence (AI). Future warfare involving manned platforms directing a larger number of proliferated unmanned systems in all operating domains cannot be realized without operator trust in combat autonomy. Today’s warfighters operate within Service cultures that tend to distrust complex autonomy and utilize autonomous systems sub-optimally in limited, supporting roles (e.g. logistics or intelligence, surveillance, and reconnaissance only).

The ACE program will increase trust in combat autonomy using human-machine collaborative dogfighting as its challenge problem, which also serves as a representation of an entry point into complex human-machine collaboration. ACE will apply existing AI technologies to the dogfight problem in experiments of increasing realism. In parallel, ACE will implement methods to measure, calibrate, increase, and predict human trust in combat autonomy performance. Finally, the program will scale the tactical application of automating a dogfight to more complex, heterogeneous, multi-aircraft, operational level simulated scenarios informed by live data, laying the groundwork for future live, campaign-level Mosaic Warfare experimentation.

Figure 1. Mosaic Warfare envisions a future in which the functions of a kill-chain are distributed across manned and unmanned assets in multiple domains

In an air domain contested by adversaries, a single human pilot increases lethality by effectively orchestrating multiple semi-autonomous, intelligent unmanned platforms from within the manned aircraft (see Figure 1). This shifts the human role from sole operator to system mission commander. In particular, ACE aims to deliver a capability that enables a pilot to attend to a broader, more global air command mission while their aircraft and teamed unmanned systems are engaged in individual tactics (see Figure 2). ACE creates a hierarchical framework for autonomy in which higher-level cognitive functions (e.g. developing an overall engagement strategy, selecting and prioritizing targets, determining best weapon or effect, etc.) may be performed by the human, but lower-level autonomy (i.e. details of aircraft maneuver and engagement tactics) is left to the autonomous system. In order for this to be possible, the pilot must be able to trust combat autonomy to conduct complex behaviors in scenarios such as the WVR dogfight.

Figure 2. In a “Mosaic” future, humans will predominantly fill the role of battle manager, requiring trust in the combat autonomy and capability provided by unmanned systems.

Human teaming within Mosaic Warfare requires collaborative interaction with calibrated levels of trust and reliance between human and machine agents. Optimally, human-machine teams form a relationship in which the human(s) in or on the loop perform only tasks which require human cognition decision-making (to include legal, moral, and ethical decisions [LME]). In these human-machine teams, the autonomous counterparts perform tasks that can be executed equivalently or better than humans while reducing risks to human life without compromising mission effectiveness or ethical considerations.

ACE will use the automated dogfight to bridge the gap from the simple physics-based automated systems currently in use to complex systems capable of effective autonomy within highly dynamic and uncertain environments at mission speeds. As shown in Figure 3, physics-based systems that comprise contemporary combat autonomy (such as autopilot-based takeoff, landing, navigation, and terrain/traffic avoidance) involve low complexity, uncertainty, and dynamism. As such, trust can be established and increased with increasing performance; however, performance is bounded, and therefore limited to simple and predictable or pre-scripted behaviors.

Figure 3. ACE will use the dogfight to bridge the gap from simple physics-based automated systems currently trusted to complex systems required in the future.

B. Program Overview

1. ACE Approach and Challenges

Within the domain of complex autonomous systems where Mosaic Warfare resides, the required operational and strategic behaviors involve high complexity and uncertainty, therefore requiring a less-bounded, highly dynamic AI capacity to execute effectively. Highly unexpected, unconventional operational behaviors may emerge from the autonomous systems that are unexpected by human operators, leading to a sense of uncertainty and lack of trust. As a result, AI for complex problem sets is largely relegated to modeling and simulation (M&S) with no satisfactory method to transition to the real world. For humans to start accepting these systems fully, trust should be established and expanded through bounded, transparent, and predictable behaviors while still preserving the potential for future novelty and situation adaptation. Only after the human-machine tactics co-evolve will the AI be encouraged to move beyond bounded, predictable behaviors, realizing its true potential for improved performance in an increasingly complex battlefield.

ACE will develop performance and trust in combat autonomy using the same bottom-up approach currently employed to build performance and trust with inexperienced human pilots. The ACE program plans to stress dogfight algorithms by applying them to missions of increasing difficulty and realism, just as is done with human pilots. The dogfight provides human pilots the opportunity to prune output behavior, enhance performance, and calibrate trust for more complicated future combat scenarios such as defensive counter air (DCA) or suppression of enemy air defenses

(SEAD).

The ACE program plans to capitalize on AI research that has recently demonstrated the ability to perform at superhuman levels in applications of increasing complexity, including gaming and self-driving vehicles. Real-time strategy (RTS) algorithms have greatly exceeded the highly publicized successes of sequential, perfect-information board games and are reaching maturity levels that make them well suited to transition to live air combat scenarios. In comparison to RTS games, WVR air combat is a relatively closed problem. While highly nonlinear in behavior, the dogfight benefits from clearly defined objectives, limited weapon/sensor integration, measurable outcomes, and physics constrained by aircraft dynamics.

This “bottom-up” approach is illustrated in Figure 4. The ACE program will move combat autonomy out of the realm of simple predictable behaviors (maneuvers) and into those of increasing complexity.

Figure 4. ACE will elevate the current state of physics-based maneuver automation into nonlinear local behaviors and then scale that capability to theater air combat.

Figure 4 illustrates the distinction between local and global behaviors with each layer relying on the ones above and below. Current combat autonomy is relegated to the lowest layer, but future Mosaic Warfare needs autonomy throughout all the layers.

The technology development on the ACE program addresses four primary challenges:

1) Increase automated air combat performance in local (individual and team tactical) behaviors

2) Build and calibrate trust in air combat local behaviors

3) Scale performance & trust to global (heterogeneous multi-aircraft) behaviors

4) Build full-scale air combat experimentation infrastructure

2. ACE Program Structure

The ACE program is constructed to address four primary technical challenges:

1) Technical Area 1: Build combat autonomy for local (individual and team tactical) behaviors

2) Technical Area 2: Build and calibrate trust in air combat local behaviors

3) Technical Area 3: Scale performance/trust to global (heterogeneous multi-aircraft) behavior

4) Technical Area 4: Build full-scale air combat experimentation infrastructure

Technical Area 1 (TA1) will be addressed under a separate BAA. Information regarding TA1 is provided in this BAA for informational and contextual purposes only. Under this BAA, DARPA is soliciting innovative proposals for Technical Areas 2-4 only. Figure 5 below shows the relationship between the different ACE program Technical Areas.

Figure 5. Relationship between the ACE Technical Areas

Performers may submit to any or all of the TAs 2-4 in response to this BAA. A proposal may only address a single TA; if a performer proposes to multiple TAs, each submission should be a separate proposal.

A government-led independent Experimentation Integration Team (EIT), facilitated by Johns Hopkins University Applied Physics Lab, will coordinate the interdependent research activities associated with the different Technical Areas. Due to the complicated integration requirements across multiple performers relying on a common experimentation infrastructure, the EIT will be responsible for the development and maintenance of the necessary Interface Control Documents (ICD) and application programming interfaces (API) associated with M&S, sub-scale, and full-scale assessments.

a) TA1: Build air combat autonomy for local (individual and team tactical) behaviors (not solicited in this BAA, for informational purposes only)

Technical Area 1 (TA1) will be solicited under a separate BAA to maximize participation of non-traditional performers. A high level description of the TA1 effort is provided here to afford context for TAs 2-4.

TA1 will develop and demonstrate WVR individual and team control algorithms with dynamic maneuvering outcomes measurable against human pilots. Implementation will begin in M&S, graduating to progressively more realistic employment including live sub-scale unmanned aerial vehicles (UAVs), and ultimately full-scale combat representative aircraft modified by the TA4 performer.

Figure 6. The TA1 performers will demonstrate WVR combat autonomy in M&S, sub-scale, then full-scale

TA1 performers will be automating pilot tactical decision-making and testing that performance against engagements of progressively increasing difficulty. For example, each stage of implementation (M&S, sub-scale, full-scale) will begin with maneuvering against an unaware opponent (e.g., cruise missile), maneuvering against a limited-aware opponent (e.g., a bomber), and finally maneuvering against an unlimited-aware opponent (e.g., another fighter aircraft). The primary success metric will be Win Probability (Pw) based on a variety of initial conditions and predetermined victory criteria derived from a notional weapon engagement zone (WEZ). Pw is computed based on the aggregate results (wins vs. loses) for a series of dogfight events. Naturally, in M&S, with faster-than-real-time and parallel play possible, the Pw statistics will benefit from a high volume of engagements, whereas the live events (sub-scale and full-scale) will have fewer engagements and thus be primarily used to validate the M&S results and enable TA2 trust assessments.

b) TA2: Build and calibrate trust for air combat local behaviors

The TA2 performer will:

Develop experimental methodology for modeling and measuring pilot trust in the dogfight combat autonomy: The TA2 performer will develop a methodology for modeling, measuring, and calibrating trust, mistrust, and distrust of combat autonomy algorithms developed in TA1 of the program. As a way to normalize quantitative measurement of trust, this methodology should include a Dual Operational Task (DOT) paradigm in which the operator is required to divide attention and workload between two, equally important but mutually exclusive tasks: a simulated Mission Commander task and the Dogfight task (see Figure 7). The experiment methodology should be designed to incentivize increased operator engagement in the Mission Commander role with success also contingent upon the dogfight engagement outcome.

Design and develop Human-Machine Interfaces (HMIs): The TA2 performer will develop interfaces that allow the pilot to interact with both the Dogfight task and the Mission Commander task. The HMI will allow the operator to execute/manage each task. For example, the interface must allow the operator to assess the combat autonomy intentions as well as take manual control to either assist or override the dogfight combat autonomy. In the case of the Mission Commander task, the intent is to provide the operator with a secondary task that must be executed simultaneously with the Dogfight.

As such, the mission commander HMI does not need to be a prototype means of doing manned/unmanned teaming (MUMT) battle management but instead representative of such an interface. The HMI will collect data on how the operator is interacting with the autonomy on each task with sufficient fidelity to conduct measurements and provide insight into the operator’s trust. It is expected that HMIs would use eye tracking (see Figure 8) or other sensing methods of measuring human attention, activity, and stress, which can be utilized to quantify and mathematically model where the operator is dividing their attention between Mission Commander and Dogfight tasks.

Model and measure pilot trust using a Dual Operational Task (DOT) implementation: The TA2 performer will conduct experiments in coordination with the EIT and the TA4 performer to implement their methodology for modeling, measuring, and calibrating trust of dogfight algorithms developed in TA1 of the program. Algorithms will be provided from each of the TA1 performers and the TA2 performer will measure pilot trust in each algorithm. Trust assessment experiments will be conducted in M&S, sub-scale, and full-scale at predefined Trust Assessment Events (TAEs) throughout the program.

Provide plan for Institutional Review Board (IRB) approval for all Human Subjects Research (HSR): It is anticipated that TA2 will involve HSR. TA2 proposals must provide a plan and draft protocol for obtaining all required approvals for any HSR.

Note that HSR is anticipated to begin in Month 17 and so any plan should also include how the performer intends to have all required approvals completed by Month 12. The plan should include a description of planned safeguards to ensure the exclusion of any personally identifiable information (PII), further detailed in Section IV.B.2.d).

Proposers are also requested to separate HSR tasks from those that do not require human-use within their Statement of Work. Detailed requirements for proposals involving Human Use (this applies to all TA2 proposals) are provided in Section IV.B.2.d).

TA2 proposals must include a suitable methodology for objective trust measurement. While reliance on an automated system (i.e., use of the automation over manual control) is relevant to trust in the automated system, it is not an adequate proxy for measuring trust. Instead, or in addition, a critical metric in assessing trust in an automated system is the frequency and timing with which an operator monitors the behaviors and performance of the automation while it is in use (with highly trusted automation being monitored infrequently, and primarily during aspects of task performance that may require human intervention).

In order to enable this modeling and building of appropriate trust, the experiment methodology should include implementation of a DOT paradigm, structured such that the operator will be responsible for dividing attention and workload between two equally important and mutually exclusive operational tasks (see Figure 7). Task 1 will be a simulation-based Mission Commander task requiring supervisory control of multiple unmanned assets. Task 2 will be a Dogfight task requiring the execution of combat maneuvers that can be optionally turned over to combat autonomy; this is a tactical engagement that is taking part within the context of the simultaneous larger mission that must be managed. The architecture of the Mission Commander simulation will be provided by the TA3 performer(s). The dogfight combat autonomy will be provided by the TA1 performers and the role of the TA2 performer is to use design-based engineering driven by the inputs from TA1 and TA3 to develop HMIs to interact with these systems simultaneously and which will be implemented by TA4.

Figure 8. The TA2 performer will measure how the operator is interacting with the combat autonomy to gain insight into trust

Figure 7. TA2 Dual Operational Task (DOT) Paradigm

Within the DOT paradigm it is up to the TA2 performer to define the data required to and mechanism of collection to conduct trust measurement. In addition to qualitative behavioral measures, performers must consider quantitative measures such as eye tracking or other methods for determining where the operator is placing attention throughout experiments (see Figure 8).

Performers may also wish to include additional metrics related to attention, workload, and stress in order to quantify and mathematically model pilot trust, and to assess the extent to which the pilot is acting as Mission Commander versus Dogfighter.

The Crosscheck Ratio (R), or the ratio of unmonitored to monitored timeshare while the dogfight combat autonomy is in use serves as a point of departure for trust measurement based on attention and workload distribution. As shown in Figure 9, dogfight combat autonomy trust calibration is a function of actual (verified) performance, Pw, and actual human trust in the system, as indicated by the Crosscheck Ratio (R). Any deviation from that curve represents a quantitative amount of miscalibrated trust and labeled as a trust calibration error (). Figure 9 provides a notional example of the trust calibration curve for an initial set of algorithms supporting automated 1v1 air combat against an unaware opponent (cruise missile), a limited opponent (bomber), and an unlimited opponent (fighter). This notional curve indicates that in the instance of an easy, unaware opponent, the Pw might be quite high, in which case having a high crosscheck ratio (greater unmonitored time to monitored time indicating greater trust in the algorithm) is appropriate. Similarly, in the case of a very difficult, unlimited opponent, the Pw might be considerably lower, and the appropriate crosscheck ratio would be lower. The TA2 performer will develop and implement a methodology consistent with the intent of the trust calibration curves described above.

The TA2 performer will also be responsible for designing the HMIs for interacting with both the Dogfight task and the Mission Commander task across all phases of the program. As described above, the Mission Commander simulation will be provided by the TA3 performer. The dogfight combat autonomy will be provided by the TA1 performers. The EIT will facilitate integration of technology developed by TA1 and TA3 performers with the HMIs. The TA2 performer will work closely with the EIT to design, develop, and integrate the HMIs for M&S and sub-scale experiments. The TA2 performer will also work closely with the TA4 performer to implement the final HMIs on the full-scale aircraft with the EIT facilitating integration.

Figure 9. Trust measurement based on ratio of time spent allowing the AI to operate unmonitored versus monitored by the pilot (R) calibrated against actual AI Win Probability (Pw)

Figure 10. The TA2 performer will develop HMIs for interaction with the Mission Commander task and the Dogfight task

A table illustrating primary ownership of the HMI effort is shown below:

Design Develop Integrate Analyze

M&S TA2 TA2 TA2 TA2

Sub-scale TA2 TA2 TA2 TA2

Full-scale TA2 TA4 TA4 TA2

Table 1. The TA2 performer owns all of the design of the HMIs and trust analysis, as well as development and integration of HMIs for M&S and sub-scale testing.

The EIT will support interaction of the TA2 performer with TA1 and TA3 performers to design HMIs for the Dogfight and Mission Commander tasks, respectively. The goal is to enable TA1 and TA3 performers to focus on algorithm development, with the TA2 performer designing and developing HMIs that support completion of both tasks within the DOT paradigm. The EIT will also facilitate necessary communications between the HMI control and display surfaces with the TA1 and TA3 software.

TA2 proposals should provide a detailed description of potential HMI and are encouraged to provide mock ups of representative visual interface designs. The TA2 performer is not restricted in the modalities that can be used (e.g., visual, auditory, haptic, etc.), and are encouraged to make a compelling case for any modalities proposed, to include considerations relevant to integration of the HMIs in M&S, sub-scale, and full-scale platforms.

TA2 proposals should provide a plan for implementing the proposed methodology for trust measurement during a series of DARPA-hosted TAEs involving operational pilots. TAEs will be 5 day events at a location to be hosted by the EIT and eventually the TA4 performer during the full-scale experiments. It is anticipated that the 5 days will consist of a combination of training and testing. Data collected during each TAE will be analyzed to develop an empirically-based trust model accounting for TA1 combat autonomy performance and human pilot interaction with the AI using the DOT paradigm.

A single award is anticipated for TA2, which will span all three Phases of the program.

Proposals to TA2 should include a Phase 1 Base period with separately priced options for Phases 2 and 3.

c) TA3: Scale performance & trust to global behaviors

The TA3 performers will:

Develop data set and model for large force exercise data analytics: Data sets for analyzing “Mosaic” tactics and strategies do not currently exist. In order to measure the performance of combat autonomy as a component in a larger effort, TA3 performers will have to develop a campaign level data set. Large force exercises currently exist which collect and retain data at a variety of disparate levels of quality and typically for reasons associated with pilot training. However, if consolidated and standardized these data sets could provide rich real-world information that would allow for both the baselining and subsequent exploitation of novel tactics and strategies. TA3 performers will develop algorithms in M&S which they will apply in a few explicit scenarios to investigate the impact of new technologies (specifically including local combat autonomy), tactics, and strategies.

Develop DOT Mission Commander Scenarios: TA3 performers will develop Mission Commander DOT scenarios for use in the TAEs. The scenarios used for global behaviors will include cruise missile defense (a specific application of defensive counter air, DCA), striker escort (a specific application of offensive counter air, OCA), suppression/destruction of enemy air defense (SEAD/DEAD), combat search and rescue (CSAR), and tactical ballistic missile launcher localization and destruction (TBM hunt). TA3 performers will work with the TA2 performer to implement these scenarios into the simulation and associated HMIs developed by the TA2 performer.

Scale local combat autonomy to, and develop battle management for, large force exercise data analytics: TA3 performers will adapt lessons learned and scale the combat autonomy developed by TA1 performers to large force exercise data analytics in M&S with the goal to assess the performance and value of combat autonomy in the context of global behaviors. The scenarios used for global behaviors will include DCA, OCA, SEAD/DEAD, CSAR, and TBM hunt.

Quantify relationship between local behavior and global behavior performance metrics: The performers will characterize the relationships between WVR combat metrics (Pw) and the metrics associated with larger Mission Commander scenarios. For example, one primary success metric for the Mission Commander scenarios will be the Kill Ratio (RK) of all blue to red forces within a given mission scenario, ideally without causing the pilot to sacrifice performance on local behaviors (the dogfight) when performed simultaneously using the DOT paradigm. As such, this will require close collaboration with the TA2 performer.

The goal of this TA is to extend algorithms to the battle management level to coordinate the actions of individual AI-driven platforms. A limited number of cognitive architectures and behaviors for individual platforms are being developed under TA1. The TA3 performers are expected to use these TA1 algorithms as part of their solution, adding both battle management level algorithms to coordinate actions between individuals and additional individual algorithms as necessary.

To support these efforts, the TA3 performers will first create a model to which large force exercise analytics can be applied. For example, a great deal of relevant progress has been made in the area of sports analytics. Leagues and teams are instrumenting their playing fields and collecting rich data sets quantifying individual and team behaviors in the context of a game. Artificial Intelligence research is being applied to these data sets to develop models of both ideal player and team behaviors, as well as discover new tactics and strategies. The case illustrated in Figure 11 provides an example in which machine learning (ML) was used to build AI “ghosts” for how an average soccer player would behave based on the player position when presented with a particular opponent tactic. Although initially developed to assist coaches and improve player performance and feedback by producing desired player behaviors for comparison, it provides tremendous potential for adversarial self-play to discover new tactics and strategies.

Figure 11. The TA3 performers will develop models to apply analytics and AI algorithms such as those being used in sports (AWS Next Gen Stats) to gain insights on tactics and strategies1

In order for this approach to have military utility, a useful model and data must exist so that the analysis may be performed. The TA3 performers will “instrument the playing field” or at the very least develop an equivalent dataset and model upon which to run large force analytics. The model will support the analysis of variables such as force composition, positioning, and employment of assets, in addition to provisioning for the introduction of new capabilities. The extent to which the model is grounded in data from recorded live large force exercises (i.e. Red Flag, Maple Flag, Mission Effectiveness Phase, and others) can increase its potential usefulness. The TA3 performers will be provided a government furnished dataset to augment performer generated/acquired data that will likely require cultivation to adapt into the simulation environment. Proposals should include details on how their model will be implemented on both the Advanced Framework for Simulation, Integration, and Modeling (AFSIM) and Next Generation Threat System (NGTS) to facilitate interaction across TAs within the ACE program. The relative merits of each simulation environment for their proposal must be explicitly described. If the performer would like to propose an alternative to AFSIM or NGTS, there must be an explanation of the expected benefit and a description of how integration with the other TAs and EIT will be addressed.

The TA3 performers will develop Mission Commander Scenarios for use in the DOT trust experimentation. There are five scenarios of increasing relative complexity to be developed under

TA3:

Cruise missile defense (defensive counter air, DCA) Striker escort (offensive counter air, OCA) Suppression/destruction of enemy air defense (SEAD/DEAD) Combat search and rescue (CSAR) Tactical ballistic missile launcher localization and destruction (TBM hunt)

The TA3 performers will design and develop a “game” that allows the operator to execute each scenario. The TA3 performers will define and instantiate the adversary posture, to include representation of platforms, capabilities, and performance. The “game” will similarly define and instantiate the Blue force and provide the operator the ability to manage Blue assets to accomplish the mission. The TA2 performer will develop the HMI for the operator to execute the Mission Commander scenarios. The TA2 performer will provide a common HMI for the TA3 performers to conform to (see Figure 12).

Figure 12. The TA3 performers will develop models informed by live large force exercises and integrate with TA2 HMIs

The TA3 performers will work with the TA2 performer to define the requirements for interaction between a pilot and battle management algorithm. This information will be used by the TA2 performer to create the HMI for implementation in the DOT to evaluate trust of TA1 AI algorithms.

The DOT will present the same Mission Commander scenario to ensure validity of the TA2 test results starting with the less complicated scenarios first. The TA3 performers will also be expected to incorporate the results of TA2 testing into their design process to preserve the custody of trust as complexity is increased.

The TA3 performers will be responsible for developing and demonstrating battle management algorithms embodied in an intelligent agent. This agent will be responsible for interacting with a human pilot who will have ultimate legal, moral, and ethical responsibility for the conduct of the engagement, and will in turn communicate commands to individual platforms. The intelligent agent may either be distributed across many platforms or may be localized on a single platform.

Blue forces under the control of the TA3 algorithms will vary, both in terms of number and capabilities of assets. Naturally, the forces available to address a particular threat will be limited and may or may not be optimally suited to address that threat. TA3 algorithms must demonstrate their flexibility by accommodating a range of blue assets, from the collective action of intelligent homogeneous platforms to heterogeneous ones working via networked environment. Performers should describe their approach to overcoming the brittleness found in many artificial intelligence techniques.

Each government-furnished Mission Commander scenario will include red forces of different levels of capability and will be delivered in order of increasing complexity. For example, the cruise missile defense mission will be delivered first with non-reactive red assets. All other missions will include non-reactive, simple, and eventually more advanced threats with the iterative introduction of more sophisticated AI. Scenario definitions will be delivered in the selected simulation environment (AFSIM or NGTS). The performer is expected to use this environment in the conduct of their work.

For the purposes of scaling local combat autonomy to battle management for large force exercise data analytics by the TA3 performers, a human operator will not manage execution of the mission.

In this instantiation of the model, the TA3 performer will demonstrate how low-level algorithms (such as those being developed in TA1 of the program) can be built up hierarchically to develop more complex mission-level capabilities. This analysis is expected to be conducted in faster than real-time to take advantage of volume that can be conducted in modeling and simulation.

The algorithms developed as a part of TA3 are expected to have the following attributes:

Scalability: Algorithms should use and extend the functionality of algorithms developed in TA1. Algorithms should also demonstrate how additional TA1-like algorithms could be incorporated in the future as new behaviors are created.

Flexibility: Algorithms should support a range of platforms and aircraft type across a range of missions. If multiple algorithms are used depending on the mission context, a selection mechanism must be developed to automatically select the appropriate algorithm.

Explainability: To the maximum extent possible, algorithms should attempt to explain decisions to the operator via the HMI since weapon employment consent and all LME will remain with the pilot.

Multiple awards are anticipated for this TA, with potential for a single performer in Phase 3.

Proposals to TA3 should include a Phase 1 Base period with separately priced options for Phases 2 and 3.

d) TA4: Build full-scale, combat-representative aircraft experimentation infrastructure

The TA4 performer will:

Supply full-scale aircraft and integrate dogfighting algorithms: The TA4 performer will implement and evaluate the algorithms developed in TA1 on full-scale aircraft.

TA4 proposals should include a description of the aircraft, ground support equipment, training and test approach necessary to implement TA1 algorithms and TA2 HMIs. If different aircraft options are offered, provide the different strategies for each option to include: the proposed airworthiness certification (e.g., FAA, DARPA), the challenges of any aircraft specific modifications, and the hourly flight rate and maximum supportable flight generation tempo for each option (i.e. cost per flight hour to maintain/operate, how many flights could be supported per week per aircraft, cost to obtain airworthiness certification). Performers will conduct an Aircraft Selection Review early (ASR) within 3 months of contract award to provide a detailed analysis justifying aircraft selection prior to proceeding with any aircraft or long-lead acquisition. Proposals should also highlight performer past experience deploying experimental aircraft to include test range coordination anticipated to facilitate the ACE program.

The TA4 performer should provide a plan for integrating TA1 algorithms into the full-scale aircraft. The performer should describe their capabilities to analyze the TA1 algorithms, embody them in real-time software for their selected aircraft, and refine/test/train the algorithms for the aircraft configuration and flight dynamics. In addition, the performer should describe their approach to ensuring the dogfight autonomy algorithms are automatically monitored so that they cannot create unsafe or unstable situations by adjusting the aircraft’s control surfaces faster than a human safety pilot can override them if necessary. The TA-4 performer will instrument their full-scale live aircraft with the equipment required to perform autonomous aircraft control and to integrate the aircraft with the instrumentation necessary to perform TAEs.

The TA4 performer will be responsible for generating and delivering a model of the full-scale aircraft for integration into the government furnished simulation environments to enable TA1 algorithm developers to tailor their algorithms to the aircraft. This will likely involve both the general support related to the aircraft and specific support related to each test event.

Develop and integrate HMIs for full-scale aircraft: The TA4 performer will coordinate with the TA2 performer to develop and integrate HMI for the trust assessment events. The HMI will allow the subject to carry out the DOT while flying the full-scale aircraft. The development and integration of the HMI will take into account the test objectives of TA2 so that the required information is both displayed on, and collected from, the HMI. The full-scale HMI should include the necessary equipment/software to provide situation assessment.

Retain safety pilot override controls and/or autopilot disconnect for trust assessments: The TA4 performer will be responsible for developing safety mechanisms on full-scale test aircraft. These mechanisms will ensure safe operation of the aircraft regardless of autonomy input and will be able to disconnect the autonomy if prohibited maneuvers are attempted.

Perform all safety/airworthiness reviews and obtain airworthiness certification for supervised live dogfight engagements: The TA4 performer is required to complete the airworthiness review process necessary to certify and approve the aircraft for flight testing in coordination with the EIT. The TA4 performer is expected to support the EIT range coordination by developing test plans and supporting test review board events.

Execute full-scale live flight experiments: The TA4 performer will provide safety pilots, adversary pilots, and aircraft in support of full-scale live flight experiments to include 1v1, 2v1, and 2v2 competitions. The EIT will be responsible for coordinating these competitions, including collecting data and scoring algorithms during test events. With the exception of the direct competitions between TA1 performers, the baseline unaware, limited-aware, and unlimited-aware adversaries are expected to be human pilots in the live competitions.

The primary objective of the TA4 performer is to develop the full-scale aircraft experimentation infrastructure capable of implementing algorithms and technologies generated by the other Technical Areas. The TA4 performer will be responsible for providing full-scale live aircraft services which will include purchase/lease, modification, airworthiness certification, training, and testing including aircraft interface development, operations, and maintenance. The TA4 performer will supply full-scale aircraft capable of integrating WVR algorithms from TA1 via the Test & Evaluation ICD maintained by the EIT. The aircraft must also integrate HMIs from TA2 and will provide the experimentation infrastructure for the TA2 trust assessments at full-scale (see Figure 13). The TA4 performer will not be responsible for any HSR. The TA4 performer will also retain safety pilot override controls and autopilot disconnect for the performance and trust assessments and will facilitate all safety including airworthiness reviews to enable supervised, live WVR engagements. Lastly, the TA4 performer will be responsible for hosting the full-scale competitions.

Modifying full-scale aircraft is complicated and will require a team with extensive experience in aircraft modification, design, and execution. DARPA is interested in a range of procurement mechanisms for TA4, from traditional delivery of aircraft as contract hardware to leasing aircraft as a service. If proposing that the aircraft be government furnished equipment (GFE), a compelling plan should include which government program office is recommended, a model for how the performer has done something similar in the past, and an estimate of any government-to-government costs. The objective is for full-scale aircraft to be as closely representative of 4th or 5th generation fighter aircraft as possible while still remaining within the time and fiscal constraints of the program. TA4 proposals are encouraged to provide multiple options with independent pricing to give the government maximum flexibility in assessing the best value to the ACE program.

Figure 13. The TA4 performer will integrate dogfight combat autonomy and HMIs onto full-scale aircraft

To enable the final team dogfighting (up to 2v2) evaluations and competition between the final two TA1 performers, a total of 4 aircraft are desired each with the dogfight algorithms installed.

However, proposals that provide multiple options with independent pricing would enable the consideration of independent 2v2 evaluations against non-AI capable or dissimilar adversaries.

Including non-AI capable or dissimilar adversary options could enable cost and schedule flexibility.

The TA4 performer will also develop and verify the pilot controls for the full-scale aircraft and HMIs to interact with local dogfight combat autonomy as well as the Mission Commander task.

Proposals should describe how the user interfaces could be integrated into the aircraft they are proposing with specific emphasis on how the airworthiness of the aircraft would be maintained in the event of an autonomy malfunction. Once selected, the TA4 performer will work closely with TA1 and TA2 performers and the EIT to ensure that the platform is compliant with the bi-directional autonomy and HMI interfaces.

The TA4 performer will NOT be responsible for development of the Mission Commander task (to be developed by the TA3 performers); however, the TA4 performer will have to work closely with the TA2 performer in order to effectively develop the HMI and integrate it into the platform. A table illustrating primary ownership of the HMI effort is shown below:

Design Develop Integrate Analyze

M&S TA2 TA2 TA2 TA2

Sub-scale TA2 TA2 TA2 TA2

Full-scale TA2 TA4 TA4 TA2

Table 2. The TA4 performer owns the development and integration of HMIs for full-scale testing.

A single award is anticipated for TA4, which will span all three Phases of the program. Proposals to TA4 should include a Phase 1 Base period with separately priced options for Phases 2 and 3.

e) Experimentation Integration Team

The ACE government-led Experimentation Integration Team (EIT) will be facilitated by The Johns Hopkins University Applied Physics Laboratory (JHU/APL), University Affiliated Research Center (UARC), and is not being solicited under the ACE TA2/3/4 BAA.

The EIT will facilitate and coordinate the interdependent research activities associated with the different TAs. Due to the complicated integration requirements across multiple performers relying on a common experimentation infrastructure, the EIT will be responsible for the management of the necessary ICDs and APIs associated with M&S, sub-scale, and full-scale experiments to include identification, development and/or maintenance as applicable. This standardization of the experimentation infrastructure is critical to assess individual TA performers. To enable independent and unbiased performer evaluation, the EIT will utilize standardized M&S environments for both the WVR engagements (TA1) as well as the larger scale, Mission

Commander Scenarios (TA3). These standardized environments will include the development of scripted adversaries for the unaware, limited-aware, and unlimited-aware opponents in all three Phases (M&S, sub-scale, and full-scale) of the ACE program. Lastly, the EIT will be responsible for the purchase, maintenance, and operation of the commercial sub-scale UAVs used throughout all three Phases of the program (and which are the focus Phase 2).

The EIT will host regular Integration Quarterlies to facilitate cross-TA interaction throughout the program; a notional layout for cross TA interaction is laid out in section I.B.3.d). The EIT will complete the following tasks in support of ACE:

Modeling & Simulation (M&S) Framework Development. Select and, if necessary, develop an M&S framework with live, virtual, and constructive components to support the development and evaluation of autonomous combat capabilities for WVR maneuvering (TA1), the evaluation of trust in those capabilities (TA2), their extension to global behaviors (TA3), and the integration into full-scale live aircraft (TA4).

1. Constructive Within Visual Range (WVR) Simulation. Provide a WVR air combat simulation environment and interfaces to maneuvering algorithms developed by the TA1 performers. Implement interface to virtual aircraft simulator in support of TA2 evaluations. Implement interface to mission-level M&S environment in support of TA3 mission command scenarios.

2. Mission-Level M&S Environment. Determine mission-level M&S environment and interface supporting TA3 algorithm development and evaluation. Implement interfaces that support integration of constructive WVR simulation instances and virtual simulators for TA2 evaluations.

3. Sub-Scale Live Aircraft. Provide sub-scale commercial UAVs in support of autonomy performance assessment and trust evaluation experiments. Implement interface to TA1 algorithms. Integrate with virtual aircraft simulator and mission-level M&S environment in support of TA2 evaluations.

4. Virtual Aircraft Simulator for HMI implementation. Provide virtual simulation environment for HSR testing to be completed by TA2 performers. Integrate with WVR and mission-level simulation environments provided in support of TA1 and TA3, respectively, and as required to meet TA2 objectives.

5. Full-Scale Live Aircraft Integration. Support airworthiness review of full-scale aircraft by the TA4 performer, DARPA, and applicable range authorities.

Adversary Autonomy Development. Develop autonomy capabilities that serve as the adversary force supporting evaluation of autonomy solutions developed by TA1 and TA3 performers at each phase of ACE.

1. Adversary WVR Maneuvering Algorithms. Provide adversary autonomy at multiple levels of performance (unopposed, limited opposed, and unlimited opposed) to assess performance of the TA1 algorithms in the constructive M&S and sub-scale live aircraft phases of the program.

2. Adversary Mission Command Algorithms. Provide adversary battle management autonomy to assess performance of the TA3 algorithms for each of the TA3 missions:

DCA, OCA, SEAD/DEAD, CSAR, and TBM hunt.

Program Evaluation Support. Support the execution of competitions between performer-developed autonomies and development of operational scenarios for TA1, TA2, and TA3 evaluations, including performer and program assessment.

1. Performer Competitions. Host competition events/demonstrations for TA1, TA2, and TA3 performers at the conclusion of each phase.

2. Scenarios and Metrics. Support the development of operational scenarios, experimental design, and selection of metrics in support of TA1, TA2, and TA3 evaluations.

3. Program Evaluation. Assist DARPA ACE team in evaluating program performers within each TA and assessing overall program success at the conclusion of each program phase.

3. ACE Program Phasing

The overall ACE program structure is shown in Figure 14. The primary focus of Phase 1 is to develop and demonstrate key capabilities in M&S. Phases 2 and 3 will implement the same in sub-scale and full-scale environments, respectively. While it is anticipated that a single performer may be used in both TA2 and TA4 ant that the government expects to carry multiple competing TA3 performers through Phase 2, the number of awards in Figure 14 is notional and subject to change at the government’s discretion.

Figure 14. ACE program schedule and phasing

a) Phase 1 (Base period, anticipated…