The Commercial-First Future of Robotics & Autonomy in Defense

There is a general consensus that the future of defense will become increasingly autonomous, and involve deploying unmanned or autonomous systems across land, air, sea, and space. It is also true that the future of all machines, including in the civilian world, will likely be increasingly autonomous – indeed, the present of consumer products and services is already beginning to be autonomous. In parts of the country right now, you can hail a Waymo, or engage FSD in your Tesla. You might also be able to order takeout and have it delivered by a Coco sidewalk delivery robot, or have a package delivered from the sky by a Zipline drone.

The state of autonomy, in on-road passenger vehicles and soon other domains, has reached the promised production-ready stage that the industry has been building towards over the last decade. The myriad technologies that enable these autonomous machines include sensor and perception technology, simulation tools, networking technology, controls, and more. Additionally, there are many related tools for developers and operators of autonomy products, such as testing, security, and fleet monitoring tools. These technologies combine to form a platform for a variety of autonomy applications, from self-driving cars to UAVs to heavy equipment for construction, mining, and agriculture.

As autonomy across all domains approaches production-ready levels for real world use cases, how can we ensure that the United States leverages our autonomy technologies to the fullest extent, and that the US warfighter is equipped with the best technologies? One part of the answer is by expanding the defense industrial base to primarily commercial companies and taking a commercial-first approach to procurement, especially in the domain of autonomy.

A commercial-first approach is compelling because of the emergence of a software-driven autonomy stack. Historically, autonomy capabilities were deeply vertically integrated and tightly-coupled across hardware, firmware, and software. In recent years, with hardware advances and tailwinds from sizable commercial investments in autonomous mobility, components like sensors and general robotics computing have become more commoditized, leaving software as the main differentiator for many instances of autonomy. In parallel, a more software-driven autonomy stack has emerged, consisting of components for perception, localization and mapping, planning and coordination and controls. These modular, reusable components of the autonomy stack allow autonomy applications to be built in a faster and cheaper manner, and serve as a platform upon which autonomy for different domains can be built. Importantly, many of these technologies are fairly use-case agnostic. While some components are tailored towards certain environments (for example, navigation systems for GNSS-denied environments being primarily useful for defense), many pieces of the autonomy stack could be used across defense and commercial sectors like construction, mining, agriculture, and more. As a result, the autonomy stack is often composed of tools and technologies that were primarily developed for more commercial use cases, but can be adapted for defense. Moreover, while there are many companies building autonomous solutions (i.e. complete systems) tailored for defense use, these companies often also leverage commercial technologies and tools in their development processes or as part of their stack, for everything from sensor integration to fleet observability. As a result, the commercial sector makes up much of the emerging autonomy stack, and should be leveraged to provide the best autonomy capabilities for American defense.

With a rapid rate of change in the dynamics of warfare, iteration speed matters. In autonomy, the best place for this kind of rapid iteration to occur is in the commercial sector. The idea of rapid product iteration – through development, testing, and continued development – is so ingrained in the methodology of early-stage technology companies at this point that it can be considered the default mode of building a startup. For many startups, their right to exist is justified primarily by speed – their ability to build, iterate, and ship fast as a nimble organization – making them an ideal place for rapid iteration in R&D. When looking at the commercial sector historically, the commercial aircraft industry and automotive sector show favorable time-to-market trends when compared to the military aircraft. Perhaps the best example in recent memory of the advantages of a commercial-first approach for iteration speed comes from Ukraine, where the country’s wartime pivot from a top-down procurement process and towards the commercial sector allowed them to outsource R&D to commercial businesses, and rapidly field commercial solutions on the battlefield. Moreover, commercial players can better design compatibility and interoperability of components when self-organizing in a bottom-up manner, compared to a top-down approach. These lessons are particularly relevant in the world of autonomy, given how much of the war is defined by autonomous systems, particularly UAS and C-UAS capabilities. Iteration speed and time to market are particularly important for the US in the current geopolitical moment, as the threat of China and a conflict in the Pacific looms – a threat that will require a robust defense industrial base and the ability to rapidly build, procure, and field advanced technologies at scale to deter.

There are also areas of frontier research in robotics and autonomy that are not ready for real-world deployment yet, but may represent promising avenues for future use across both the defense and commercial sectors. These include areas like fine-grain manipulation and locomotion capabilities for dextrous and legged robots (including humanoids and quadrupeds), end-to-end robot learning, and the promise of ‘robotics foundation models’ that might allow autonomous robots to generalize across tasks, environments, and embodiments. These advances in autonomous robotic capabilities can have enormous potential for applications in defense, but at present are primarily research problems, often focused on household or industrial tasks. The costs of this R&D are best borne by the commercial sector, and companies can develop novel technologies on their own dime with clear market signals from the defense sector. The history of self-driving demonstrates how this collaboration across defense and research can work – the DARPA Grand Challenge set the direction for much of the field of autonomy as a prize competition for autonomous vehicle research, and teams from various universities developed vehicles to compete in the challenge. In the current era, where many leading AI companies look more like research labs, much of the cutting edge robotics and autonomy research is being done by companies. The early productized technology in novel areas of research in robotics and autonomy is thus likely to come out of the commercial sector, which can then be adapted for defense applications.

Moreover, leveraging commercial technologies in the autonomy stack is a positive-sum strategy for the entire defense industry. As many commercial companies have built use-case agnostic products for the autonomy stack, defense-first developers of applications and systems integrators can leverage these tools in their own solutions, adapting commercial technologies to the defense environment. The commercial autonomy stack lets defense-first companies build better capabilities and products, can help lower time-to-market by providing modular autonomy components, and ensures that the defense industrial base provides the warfighter with the best autonomous technologies.

As the future of warfare continues to become more autonomous, the defense industrial base must leverage the autonomy capabilities developed by America’s technology industry to the fullest extent. Operationalizing these changes means taking a commercial-first approach to procurement, that focuses on enabling commercial companies to deliver superior solutions at lower costs and faster speeds. Such an approach can ensure the American warfighter is equipped with the best technologies for the battlefield, allow cutting edge autonomy products to be fielded faster, enable more cost-efficient R&D for the long term, and strengthen the American defense industry as a whole.