OPEN TO MEMBERS
The ARCNet Consortium is a collaboration of industry and academic members working in concert to expedite the development of technology and to address current and future Air Force technological needs. Membership in ARCNet is required for access to full opportunity details and bid submission. Interested in learning more about ARCNet? Click here.
SAR Sensor Development and Integration for GPS-denied Navigation and Target Acquisition R&D Efforts October 8, 2020
The Air Force Research Laboratory, Munitions Directorate, is seeking proposals for a radio frequency (RF) sensor appropriate for Synthetic Aperture Radar (SAR) and other radar mode processing on an unmanned aerial vehicle (UAV). This sensor acquisition effort is intended to remove GPS dependencies associated with the operation of an RF sensor and to publish raw or nearly raw sensor data in support of real-time, onboard algorithmic development and flight testing of non-GPS navigation and other signal processing concepts.
Mobile Unmanned/Manned Distributed Lethality Aerial Network (MUDLAN) provides end-to-end network architectures for connectivity between multiple airborne, surface, and ground platforms, with additional technologies to produce a larger Intelligence, Surveillance & Reconnaissance (ISR) networked capability, linking multi bands such as Ku, X, W, Tactical Targeting Network Technology (TTNT), and Link 16. This project will develop a network architecture for an operational demonstration including a path for ALT PNT integration.
In order to plan for FY21 scoping discussions, TRMC’s ADAS project requests information from the ARCNet consortium members for ideas of addressing T&E of autonomy systems, with a specific emphasis on Natural Language Processing (NLP) across a program’s life cycle. Recent technological advancements in the area of NLP have produced open source and commercial off the shelf (COTS) products that are transforming the landscape of information analysis across industries. NLP tools such as GPT-3 and BERT are just a couple of recent examples of AI models/services that could assist with expediting and scaling automated tools to support workflow processes that are dependent upon language-based knowledge products. The objective here is to explore what are the opportunities whereby NLP could be used to transform and accelerate traditional manual processes typically performed by humans. Specifically for ADAS, how can we apply NLP to support the unique challenges of an autonomous system that is responsible for making decisions in the context of a warfighter mission or business process ISO fielding, testing and delivering autonomy systems? How can we leverage NLP to support the life cycle with emphasis on RDT&E workflow processes?
The goal of the “Link Formal Models to Implementation” (LFMI) ADAS subtask is to develop new methods and tools for demonstrating that an autonomous system implementation satisfies its formal model(s). Although a formal model that specifies the required behavior of an autonomous system is valuable, the ultimate objective of the development process is to produce an autonomous system that satisfies its requirements. To detect defects in a formal model, such as inconsistencies and violations of system properties, one can use static analysis tools, such as consistency checkers, SMT solvers, and model checkers. To detect defects in the implementation, in contrast, one can use dynamic analysis tools that check the implementation during run-time. One effective approach to evaluating the correctness of an autonomous system’s implementation is to apply black-box testing to determine whether the implementation, given a sequence of system inputs, produces the correct system outputs.
The ASSURED DEVELOPMENT & OPERATION OF AUTONOMOUS SYSTEMS (ADAS) research project is developing new capabilities for testing and evaluating (T&E), and verifying and validating (V&V) autonomous systems. In ADAS, a joint Army/Air Force/Navy team is developing an integrated suite of models, methods, and tools for building prototype autonomous systems that are assured to satisfy critical safety and functional requirements. This effort will support the “Link Formal Model to Implementation” (LFMI) ADAS sub-project team in developing and evaluating new methods and tools for demonstrating that the implementations of run-time monitors and run-time enforcers (such as watchdogs) satisfy their requirements—in particular, that they detect and avoid hazards that could lead an autonomous system to behave unsafely. This effort will support the LFMI sub-project team by assisting in the development of formal models of run-time monitors and run-time enforcers, by helping to formally verify the formal models, and by helping provide high assurance that the run-time behavior of code implementing run-time monitors and run-time enforcers behaves safely and correctly.
The Air Force Research Laboratory, Munitions Directorate (AFRL/RW) is interested in supporting the continued development of factor graph methods for real-time navigation state estimation of collaborative munitions with limited access to Global Positioning System signals. Inter-vehicle range and/or bearing measurements have been shown to reduce the drift of the onboard inertial navigation systems, and although factor graphs provide a useful framework for the estimation problem, some challenges still need to be addressed. Since navigation accuracy for this scenario has also been shown to depend on the vehicle trajectories, AFRL/RW is also interested in path planning methods that minimize final position uncertainty and thus improve mission effectiveness.
Autonomy & Artificial Intelligence DevSecOps Test & Evaluation SW Developer Services – Transition to Iron Bank (ADAS- T2IB) July 21, 2020
The intent of this project is to identify an existing government-owned & licensed “Autonomy and or Artificial Intelligence” (AAI) Software (SW) DevSecOps Test & Evaluation Tool/Service (aka., AAI SW DevSecOps Application Test Harness), and convert this legacy application/service into a containerized cloud managed service(s) to be used by the department’s software developer community to automatically test the performance, behavior and capabilities of their new AI algorithms and autonomous capabilities ISO the joint warfighter. This is a pathfinder project, whereby all future AAI SW T&E tools & services will be managed and delivered as containerized services as part of the DoD’s distributed managed service strategy on cloud. ARCNet respondents will identify and recommend an existing legacy AAI SW DevSecOps Tool/Application, and then convert this legacy application/service into a containerized, production-ready SW T&E service, to be delivered on the DoD PaaS repository known as the DoD’s PlatformOne (PartyBus, IronBank, etc.).
The use of autonomy requires novel approaches for teaming human operators and machine partners. Decades of research have shown that the introduction of automation can drastically change the demands of human operators morphing humans into monitors of the technology (Parasuraman & Riley, 1997). In doing so, humans are faced with novel challenges of increased workload, mode confusion, and miscalibrated trust which can hinder performance (Hoff & Bashir, 2015). Contemporary approaches seek to blend humans and machines into seamless human-autonomy teams (HATs), which raises new challenges for both the approach toward autonomy and the human-machine interfaces required to facilitate shared awareness and shared intent between the humans and machines. Given the dearth of science on what approaches are most effective for HATs, the development of interfaces to support manned-unmanned teaming (MUMT) is as much art as science. Thus, research is needed to understand the various approaches to MUMT, compare them based on the extant HAT literature, and establish guidelines for effective MUMT for future DoD systems.
AFRL is interested in new methods for planning of multiple air vehicles conducting sensor data collection under the threat of attrition. Approaches must balance the collection and return of sensor data with the loss of assets, yielding a multi-objective optimization problem with probabilistic uncertainty. Additionally, formal bounds on performance and complexity are needed to provide assurance of solution quality and computation time across a wide variety of potential problem instances.
Swarms of opposing autonomous aerial vehicles imply radical new concepts in tactics and behaviors. The focus of this work will be to identify strategies and develop algorithms to analyze a swarm’s behaviors and use this data to counter its tactics. Strategies of interest include decentralized and cooperative guidance laws, weapon target assignment, simultaneous arrival at location(s), formation control, and a Battle State Estimation Engine.
In the future, Air Force (AF) Unmanned Aerial Vehicles (UAVs) will need to perform missions more autonomously while in contested environments. Having onboard, mission-relevant “knowledge” loaded before mission start will enable the UAVs to interpret incoming mission sensor data in context and provide better situational awareness which will in turn lead to better decisions. This will be true both on an individual platform basis and for a collective of collaborating UAVs. The goal of this project is to explore the concept of an onboard world model/data store that utilizes a synergistic graph-based and machine learning approach to provide a priori mission knowledge that is constantly updated with incoming knowledge from exploited mission sensor data. The project will develop concepts and implement one or more representative instantiations of a world model to show and test capability.
Autonomy & Artificial Intelligence – Next Generation – Live Virtual Constructive Test Range June 11, 2020
The vision of this next generation range is to accelerate innovation and invention of new tactics, techniques and procedures (TTPs) leveraging manned and unmanned test (MUMT) systems in support of the armed services and joint commander(s). Enabling warfighters at all levels to understand how humans and machines will operate, together, in an operational setting, while ensuring they are safe and ethical. This test range will be like no other range ever built in the department. It will be a range that is architected as a live, virtual, and constructive (LVC) test environment with scalable compute, storage (cloud) and network resources (5G), where elements of a live physical world are blended with those of a virtual world, through “Modeling & Simulation-as-a-Service” (M&SaaS), constructing a scalable, dynamic, yet safe environment for MUMT experimentation and testing. The constructive ability includes environment and event composition as well as configurable instrumentation to harness and evaluate resulting data and operations from the edge.
The Office of the Under Secretary of Defense for Research & Engineering (OUSD – R&E), Assured Development and Operation of Autonomous Systems (ADAS) Project, in partnership with US Army Futures Command (AFC), the DoD’s Joint Artificial Intelligence Center (JAIC), and the ARCNet Consortium, solicits proposals to develop a next generation collaboration platform for the DoD’s data scientist workforce that supports autonomy and artificial intelligence projects and programs. This project is known as “Coeus”, named after the ancient Greek god of “query, questioning” &”intelligence”. This workforce includes data scientists, data analysts, data labelers and machine learning (autonomous & artificial intelligence) engineers & developers. The purpose of Coeus is to make data, DevSecOps, software and infrastructure resources easy to access and share in collaborative settings to maximize reuse and development synergy supporting Autonomy and Artificial Intelligence (AAI) AAI data projects. Additionally, the platform enables a continuous integration / development lifecycle supporting a larger software development process across DoD. Coeus will be developed as a DoD wide Software-as-a-Service (SaaS) service supporting the department’s data scientist community. It will be developed with software tools, services and infrastructure provided by the DoD’s “Platform One” and “Cloud One” Program. Coeus will be developed as a containerized, cloud native SaaS service. This maximizes Coeus’s flexibility to run on different DoD cloud PaaS environments to support the software development community across the department.
Time Critical Targeting in Urban Environments (TCTUE) focuses on developing a targeting system to enable multiple unmanned combat air vehicles (UCAVs) to search for and track a mobile target at night or in weather in any known complex urban environment. The overall objective of TCTUE is to demonstrate the impact of collaboration using three UCAVs with on board sensors, supervised by a single human operator, to find, fix, target, and track a single target vehicle in a cluttered known urban environment. In addition, TCTUE should provide cross-swarm intelligence during all mission phases to: minimize the time to correctly identify the target in clutter, minimize the time for the operator to decide to launch a weapon, and provide the operator the ability to send in-flight target updates to the weapon following handoff from the UCAV. Finally, TCTUE will dynamically generate a flyable trajectory for a weapon through an occluded and congested urban scene. The TCTUE system should be demonstrated in software, hardware in the loop, and in live flight on multiple UAVs, in collaboration with the AFRL Munitions Directorate flight team.
Cognitive Support for Assessing XAI (eXplainable Artificial Intelligence) Systems (CSAXS) March 27, 2020
Across operational environments there is a growing urgency to empower decision makers and analysts to capitalize on a range of rapidly-evolving “analytics” capabilities, including decision aids, algorithms, automation, autonomy, and AI/ML. The 711 HPW seeks to exert intellectual leadership in the development of AI systems that are learnable, useable, useful, and understandable. The objective of this effort is to develop a suite of theory-driven tools that will increase the likelihood that these technologies will not only be fielded but will measurably enhance warfighter performance. The XAI assessment framework is intended to be releasable to the data analytics research and development community to aid in the creation of mission-effective AI/ML systems across the DoD.
Next generations of military aircraft require dramatically increased propulsion, power, and thermal capabilities without corresponding growth in system size and weight. As a result, right-sized, efficient physical architectures that robustly ensure safety-critical functions are crucial for warfighter efficacy. However, such architectures exhibit reduced safety margins, hybrid dynamics, and increased complexity, which prove challenging to develop, verify, and validate. The goal of this project is to develop tools amenable for showing such cyber-physical systems safely and securely deliver increased next-generation propulsion, power, and thermal capabilities.
The Air Force Research Laboratory, Munitions Directorate, Weapon Dynamics, Guidance, Navigation & Control Branch (AFRL/RWWN), is seeking proposals to mature a collaborative relative navigation concept for distributed vehicles. This low-bandwidth whole-navigation method has been demonstrated in both software and hardware, enabling GPS-denied/degraded vehicles to cooperatively share information to improve localization accuracy. AFRL/RWWN’s goal for this effort is to mature and demonstrate the relative navigation concept using mature software and tactically relevant hardware through a series of increasingly complex flight tests.
AFRL is interested in developing and maintaining a simulation environment built specifically for the autonomous flight domain with integrated machine learning (ML) support. AFRL is seeking interested parties to submit proposals for software solutions to address this need.
Manned-unmanned teaming operations require seamless operation across all domains of warfare. For Group 4-5 UAS these operations include all phases of flight from takeoff and departure in a terminal area environment through tactical mission operations as well as arrival back to a terminal area and recovery. Autonomous technologies are required to seamlessly integrate into the airspace and securely coordinate with Air Traffic/Air Combat Controllers to increase safety and share situation awareness for nominal and contingency operations.
As systems exhibit increasing levels of autonomy, the overall verification and validation burden has increased. Assurance cases address this challenge by presenting an argument for the system’s appropriateness for its intended environment as well as a forum for tracking the various heterogeneous evidences that support the argument.
The Air Force Research Laboratory’s Autonomy Capability Team 3 (ACT3) is chartered to operationalize AI at scale through the development and application of an AI software platform. ACT3 needs to provide advanced applications that utilize autonomy and AI technologies against various Air Force problem sets. Given the need to re-use underlying technologies, a suite of base capabilities are required, with an initial focus towards specific use cases.