Trusted autonomy development across all domains
How can the DoD accelerate trusted autonomy development across air, land, and sea?
Welcome back to the Nexus Newsletter. In this edition, we preview an upcoming webinar with AFWERX, share information about our booth and happy hour at AUSA, highlight a recent production contract to deliver synthetic data for AiTR, and share a recent podcast appearance focused on building trust in autonomous systems.
Nexus webinar: Aerial autonomy
There is a massive demand signal for autonomous aerial systems. How can the Department of Defense accelerate the development, testing, and deployment of those capabilities?
Lt. Col Bryan Ralston, Autonomy Prime Lead at AFWERX, recently stopped by our office to discuss:
Where autonomy fits into the Air Force’s Operational Imperatives
The status of Air Force autonomy efforts and the stakeholders involved
Efforts to accelerate the test and evaluation of autonomous systems
AFWERX Autonomy Prime’s role and priority focus areas
Why Phase III SBIRs are an important success metric for AFWERX
The distinction between capabilities and platforms
Tune in to the conversation on Thursday, October 5, 2023, at 1:00 PM ET. Can’t attend live? Register and we’ll send you a link to the webinar after it airs.
Meet us at AUSA
AUSA is next week. Stop by booth #2955 to meet the team and learn how we’re accelerating the development and deployment of autonomous systems for defense customers in all domains.
Also, don’t forget to register for happy hour! Join us on October 10 at 5:00 PM to network with other autonomy experts and exchange ideas with the team accelerating autonomy development for the Army’s major robotics programs. Drinks and food will be provided.
On contract: Delivering synthetic data for AiTR
We recently announced that Applied was selected for a production contract to deliver synthetic data to train an aided target recognition (AiTR) algorithm to identify and track adversary systems.
“Aided target recognition is fundamentally a perception problem - something that we’re very familiar with from our work in the commercial autonomous vehicle space. The performance of a machine learning model that lies at the core of a perception system depends on the quality and quantity of labeled training data that’s made available to it,” said Peter Ludwig, Co-Founder and CTO of Applied Intuition. “When it comes to military applications of perception systems, including AiTR algorithms, real-world data is often impossible to collect in the quantities needed to enable success. To remedy this problem, autonomy development teams leverage high-fidelity, physically accurate synthetic data with pixel-level annotations to rapidly generate training data and ensure that models are able to accurately and reliably identify targets, objects, and obstacles.”
Building trust in autonomy
Ahmed Humayun, Head of Federal Growth at Applied, recently joined Second Front’s “All Quiet on the Second Front” podcast to discuss the importance of building warfighter trust in autonomous systems, leveraging commercial innovation for government applications, rethinking the requirements development process to improve outcomes, and more.
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:
The War Zone | Russian Su-34 Strike-Fighter Seen Covered In Tires
Key quote: "A covering of tires could well be calculated to break up the infrared signature of these aircraft, to confuse cruise missiles using image matching for targeting. This technique is also frequently referred to as DSMAC (Digital Scene Matching Area Correlator) or ATR (Automated Target Recognition) when used in cruise missiles. As we have previously discussed, using DSMAC/ATR would provide land attack versions of Ukraine’s homegrown Neptune missiles with a significant advantage, making them largely immune to electronic warfare jamming. At the same time, their approach to the target would not involve any telltale radio-frequency emissions, thanks to the passive nature of the targeting."
The tires are now also known to degrade space-based synthetic aperture radar (SAR) surveillance that can look right through clouds, providing all weather targeting day and night for standoff strikes. The tires could similarly make it harder for weapons with radar seekers to find their targets, although we are not aware of a Ukrainian standoff munition that uses this concept of terminal guidance. Still, that hasn't stopped Russia from deploying simple countermeasures for such systems before. In fact, they have done exactly that on a relatively wide scale.
Our take: Russian efforts to obscure fighters and bombers on runways with tires, likely to disrupted automated target recognition systems, raises an important question: How can you quickly train algorithms in response to battlefield conditions? In short: Synthetic data. High-fidelity, physically-accurate synthetic data enables development teams to rapidly generate new training data based on observations of a dynamic battlefield. Development teams could rapidly generate synthetic scenarios that accurately mimic occlusions on the battlefield in order to train perception systems to identify targets through changing camouflage.
Defense One | Navy’s 2-year-old robot task force eyes more AI
Key quote: For his part, Clark said he wished the Navy had skipped 4th Fleet and brought unmanned experimentation straight to the Pacific Fleet, where environmental conditions and the demands on drones are different. Pacific applications will likely focus less on surveillance than effects: “targeting, combat identification, and then kinetic or electronic warfare maybe,” he said, citing conversations with Pacific Fleet officials for his recent study on unmanned systems.
Clark said this will likely mean moving away from TF59’s contractor-owned model.
“One of the problems that 5th Fleet identified was, if we want to take this model and apply it to military missions, we'd have to probably have them be government-owned,” he said.
In the next few years, Clark said, he hopes to see a “bifurcation” of TF59’s work, with ISR missions shifting to regional partners and the Combined Maritime Force, and the Navy taking on transforming the drones for military operations.
Our take: Moving from TF59’s contractor-owned, contractor-operated model to a more traditional, government-owned model in the Pacific is an interesting development. For TF59, the COCO model allowed for rapid acquisitions of commercial technologies and iterative improvements in the field. It will be interesting to see whether unmanned systems are deployed in the Pacific at speed, and to what degree those systems are rapidly improved, in the more traditional government-owned model.
Defense News | Defense Innovation Unit seeks modular test system to scale drone tech
Key quote: DIU’s new solicitation emphasizes the Pentagon’s need for an open-architecture platform that can test, integrate and qualify a range of subsystems and materials without the “exquisite components” and “labor-intensive manufacturing” that slow down existing production processes.
“The Department of Defense replenishment rates for unmanned aerial delivery vehicles are neither capable of meeting surge demand nor achieving affordable mass,” the notice states. “Narrow supply chains, proprietary data and locked designs result in a lengthy timeline to transition new technology into usable capability and limit production and replenishment rates.”
Our take: This is an important effort. With that said, it is essential for this effort to also incorporate software development, integration, and testing from the beginning. A software development and testing pipeline that draws on best practices from the commercial autonomous vehicles space is critical in order to respond to emerging battlefield requirements and employ capabilities at speed. Without the capacity to iteratively develop, test, and field autonomy software, any enterprise test vehicle effort will be doomed to quick obsolescence.