Autonomy at speed: Actions for Congress & the Air Force
Steps to accelerate autonomy development.
Welcome back to the Nexus Newsletter. In this edition, we highlight two letters calling on Congressional leaders to fund and support initiatives that would accelerate defense autonomy, share key insights and recommendations from a recent report on Air Force approaches to AI T&E, and comment on recent news, including the upcoming National Defense Industrial Strategy.
Appropriators: You can accelerate autonomy
Congress, including Appropriators, has a critical role to play in accelerating the development and deployment of trusted, all-domain autonomous systems that will promote U.S. competitiveness, enhance deterrence, and improve warfighter effectiveness.
That’s why we were proud to contribute to two letters sent by the Association for Uncrewed Vehicle Systems International (AUVSI) and the Silicon Valley Defense Group (SVDG) to the leadership of the Senate and House Appropriations Committees. The letters call on Congress to support specific initiatives in FY24 that are designed to improve the Department of Defense’s ability to develop, test, and deploy autonomous systems at scale.
The letters call for:
The creation and full funding of an enterprise development and testing platform for autonomous systems. The AUVSI letter urges Congress to fund the AEP at $50M, which will enable the CDAO to accelerate autonomy development, testing, validation, and deployment, based on best practices from the commercial autonomy industry.
Restored funding for leading defense ground autonomy programs included under the Tactical Unmanned Ground Vehicle (TUGV) line item in the defense budget, including the RCV, OMFV, and CTT programs. The AUVSI letter requests the restoration of $19.6M in funding cut from the TUGV program to match the funding requested by the House and President’s budget.
Support, authorities, elevated reporting structure, responsibilities, and increased funding for the Defense Innovation Unit (DIU). The AUVSI and SVDG letters both highlight how DIU has demonstrated effectiveness in serving as the connective tissue between the innovation engine of Silicon Valley and the national security acquisition ecosystem, and urge Congressional leaders to codify DIU’s elevated reporting structure, expanded responsibilities and authorities, and support the creation (and funding) of a $1B “hedge portfolio” that will enable DIU to accelerate the fielding of innovative capabilities to the services and combatant commands at the speed of relevance.
We are proud to join other leading technology companies, investors, and associations in urging Congress to support and fund key initiatives that will help the Department of Defense develop, test, and deliver critical capabilities.
USAF T&E of AI-Enabled Systems
In September, the National Academies of Sciences, Engineering, and Medicine published a report examining the technology landscape, outlining the challenges, and recommending a roadmap to improve the Department of the Air Force’s (DAF) ability to develop, test, validate, and deploy AI and autonomous systems. The report was developed at the request of the DAF and includes a discussion of how the Air Force’s traditional approach to T&E is insufficient for AI-enabled systems. It then recommends a DevSecOps approach to AI T&E.
Key points:
The traditionally strong separation of DT&E and OT&E for weapon systems are not workable when it comes to evaluating the performance and safety of AI-enabled and autonomous systems.
AI must adapt to stay relevant, and therefore it requires agile development methodologies that integrate D/OT&E directly into the development cycle.
As it stands now, the DAF is “currently not prepared for this level of continuous integration and continuous deployment or delivery.”
Autonomous vehicle (AV) development, testing, and deployment is the modern example of AI being integrated into safety-critical systems that require complex system-level integration, and serves as a valid case study for the Air Force due to similarities to Air Force autonomy program goals.
An iterative, DevSecOps approach to AI T&E that enables continuous improvement post-deployment must be able to function in a decentralized environment and allow for model maintenance and retraining while in the field: “The DAF should invest in synthetic data engines, live virtual constructive environments, data repositories, and support for digital twins representative of their modalities and platforms of interest to facilitate rapid model retraining and maintenance.”
On the last point, the report states: “The decentralized nature of DAF operations means training cannot be supported by standard commercial offerings. The committee knows of no commercial off-the-shelf solution [that] presently supports these requirements.”
Our take: The DAF’s future operating concept, including Agile Combat Employment (ACE), means that the future development methodologies must be able to maximize the utilization of multi-domain data that is specific to the unique area of operation to provide both feedback and training opportunities. The iterative development, training, evaluation, and employment of perception models should be focused on localized pipelines to maximally serve the unique battlefield requirements detecting, for example, a BMP-3 in Ukraine versus Iraq. There IS a DevSecOps autonomy development, testing, and validation pipeline rooted in the autonomous vehicle industry that enables continuous improvement over the entire lifecycle of a program, while being platform agnostic and optimized to perform across decentralized teams: the Applied Development Platform (ADP). ADP makes it easy for development teams - including lab-based teams and forward-deployed engineers - to quickly ingest and visualize sensor log data to understand system performance, create curated training datasets made up of real-world and synthetic data, evaluate model improvements in sensor and object-level simulations, and deploy improved models back to platforms. Not only is there an off-the-shelf solution ready to be utilized by the DAF, but this solution and proven pipeline can be employed to increase the speed of deployment while reducing the program cost in contrast to the DAFs existing approach.
News we’re reading
Autonomy is everywhere in defense these days. Make sense of the latest headlines by reading key quotes from recent articles of interest, plus brief commentary from Applied Intuition’s government team, below:
Foreign Affairs | AI Is Already at War
Key quote: To do so, the Defense Department and the intelligence community will have to invest more in accelerating AI adoption. They can start by building common digital infrastructure systems that share the same standards to ensure interoperability. The infrastructure would include cloud-based technologies and services; common data standards; validated data sets; shared access to secure software stacks; sophisticated tools for the testing, evaluation, and validation of AI models; and secure application programming interfaces that control who gets access to what information at various levels of classification. The goal would be to give developers the data, algorithms, tools, and compute power—or high-speed computing power—they need to create, test, validate, and use new AI tools.
Our take: The former Under Secretary of Defense for Policy is spot on with this recommendation, which is only one of several (spanning everything from acquisition policy to workforce development) that she outlines in this piece. Investing in the enterprise-wide development and testing infrastructure for AI-enabled capabilities, rather than solutions stovepiped within a single program or service, is a surefire way to accelerate development and, ultimately, adoption. The AUVSI letter we recently endorsed calls for the creation of this type of enterprise-wide development and testing platform for autonomy, specifically.
Breaking Defense | Pentagon’s first industrial base strategy meant to ‘catalyze generational change’
Key quote: There are four key areas the strategy focuses on: having resilient supply chains, workforce readiness, flexible acquisitions and a focus on economic deterrence and economic security.
“While these four priorities are designed as a department-wide industrial strategy for how the DoD will build a modern industrial ecosystem, of course, the department cannot do this alone,” she said. “We are engaged with others to share our aim of a future with a robust modernized industrial base. This includes and very much, I believe, both traditional and more innovative or non-traditional suppliers.”
Our take: We’re looking forward to seeing how the industrial base strategy addresses and incorporates the role of non-traditional suppliers. Much of the conversation around the industrial base has - and for good reason - focused on the capacity to produce adequate volumes of artillery shells, armored vehicles, ships, and aircraft to support conflicts in Eastern Europe and the Middle East without negatively impacting readiness of our own armed forces. While that is critical, today’s defense industrial base extends far beyond hardware and munitions manufacturing, and the supply chains that support them: A robust defense industrial base strategy must place a significant focus on the software side of the equation.
Defense News | Improper storage damaging $1.8 billion in Army ground combat equipment
Key quote: The spare parts stored at these facilities support some of the Army’s flagship land fighting technology, including the Bradley Fighting Vehicle, the Stryker Armored Combat Vehicle, and the Abrams tank, the report noted.
Two-thirds of the inspected parts — valued at $1.31 billion — exhibited “critical” deficiencies, meaning they were “deteriorating and in immediate danger of moving to a lower condition code,” a measure of utility. In all, investigators determined that $1.8 billion (92%) of the inspected equipment was crumbling or at “increased risk” of falling apart because DLA storage facilities failed to follow Pentagon standards.
Our take: This cautionary tale of hardware sustainment holds lessons learned for Army modernization programs focused on developing increasingly software-defined next-generation combat vehicles. Programs must be proactively constructed in a manner that enables ongoing, iterative improvements to vehicle software over the entire lifecycle of the program, or risk having outdated, neglected, and irrelevant algorithms that negatively impact readiness and effectiveness.