Securing Competitive Edge: How the U.S. Military Can Benefit From a Software-Forward Approach
Apr 5, 2024
.webp)
Autonomous behaviors and human-machine teaming in military platforms will produce a technological edge that determines the outcome of future conflicts. However, the U.S. military still approaches platform architecture with processes built for the industrial age. Future defense programs will struggle to keep pace with the AI revolution if they are built with traditional, hardware-centric designs. As the U.S. military presses ahead into the digital age, program executive officers can leverage hard-won lessons from top vehicle manufacturers that offer a new vision for software-defined vehicles to secure America’s competitive edge.
Traditionally, the factors differentiating automotive products were hardware components—the body style, engine design, or powertrain specifications. In the future, the best vehicles will instead be distinguished by the quality of onboard and offboard software. Success in the fiercely competitive automotive market will hinge on delivering new features that consumers have come to expect like autonomous driving and access to cloud-based services.
The concept of a software-defined vehicle emerged to address a problem plaguing even the best automotive manufacturers pursuing this vision. Premium vehicles today have as many as 100 onboard computers, known as electronic control units (ECUs), that govern every function of the vehicle from infotainment to emergency braking. Onboard software has grown to a point where the average modern vehicle has roughly 50-80 million lines of computer code. To further complicate matters, vehicle manufacturers may source these parts from as many as 20-30 different suppliers making integration and compatibility a nightmare.
The physical design connecting these components has become correspondingly complex. New vehicle capabilities necessitate a vast array of on-board sensors like lidar, radar, and cameras each with unique data processing requirements. As new components are added, integration complexity increases exponentially making comprehensive verification and validation (V&V) testing a challenge. Root cause analysis to find and patch bugs has become equally difficult. To anyone involved with the F-35 program over the past two decades, these problems may start to sound familiar.
Another problem facing auto manufacturers is that maintaining these features at the edge now requires them to continuously update onboard software over a vehicle’s lifetime. In the same way that we now expect new features with every update of our smartphone’s operating system, drivers will come to continuously expect new functionality from their cars. This might include connectivity to cloud-based services or even other vehicles. For any AI-centric technology, developers must have the ability to backhaul relevant data from their operational fleet over-the-air (OTA) to label, train, and recursively improve machine learning models.
Participants in the defense innovation ecosystem will start to see the parallels emerge between the vision the commercial auto industry is working toward and the critical technologies the Department of Defense seeks to implement. These capabilities will emerge in rapid, iterative development cycles driven by experience in testing and operational deployments. All of these depend on the ability to rapidly build, test, and deploy software. They will require platforms with largely static hardware configurations to accept constant updates over their lifetime, something the U.S. military has thus far struggled to do.
The solution is to fundamentally reimagine how a vehicle is designed. Defense programs should consider bifurcating their software and hardware development into two separate processes. This decoupling allows software to iterate at its naturally higher frequency when compared to hardware design. Both of these development efforts then occur in parallel, linked through a common set of defined program requirements. These requirements drive integration testing across the entire lifecycle of the program to ensure the two do not diverge.
In practice, this development occurs as a collection of many distinct processes that can be logically grouped into two categories. The first is what happens on the vehicle. These decisions are partially about simplifying the design of the vehicle itself through best practices like mandating common sets of flexible application programming interfaces (APIs) to abstract away from proprietary hardware functions. The use of a few powerful centralized computers to supervise the dozens of independent ECUs will reduce cost and help to ‘future-proof’ a system.
%20(1).webp)
Powerful compute also provides the ability to conduct on-board data logging and triage. This is essential for root cause analysis in failure modes, predictive maintenance, and for gathering training data to improve onboard AI/ML models. All onboard vehicle software modules should be containerized to reduce dependencies and improve the reliability. At the highest level, onboard software will be responsible for providing the user interface and user experience (UI/UX) to the user operating the vehicle. A future military platform might one day link to an app store where tank drivers or fighter pilots download and customize the functionality of their platform to their unique preferences.
Equally important are the processes that happen off the vehicle. Testing progressions with model-in-the-loop (MiL), software-in-the loop (SiL), and hardware-in-the-loop (HiL) approaches can help accelerate the materiel release process for any software update. These techniques exercise a large set of unit, integration, and system-level tests that provide traceability of requirements to code, enabling the successful deployment of software in a modern continuous integration, continuous delivery (CI/CD) pipeline. Programs will also require tools for V&V testing to produce and track performance metrics from these tests over time. This is critical to ensuring that a software patch or release will not adversely affect a deployed system.
During operations, wireless data connections to the platform must enable remote monitoring of the health and status of vehicles in the fleet. For any onboard modules that employ AI/ML models, real-world deployments provide an essential (and in some cases potentially the only) source of valuable training data. Data scientists supporting military programs must be able to task deployed platforms in the field to collect against emerging requirements and backhaul this data to the point of processing. Connectivity also enables data collection for critical fleet management functions such as predictive maintenance.
Military platforms were once similarly made distinct by their hardware characteristics like armor thickness, main gun caliber, or tread pattern. The future will be very different. Instead, it will be the ability of military platforms to sense their environment, behave autonomously, and act cooperatively that will give one side supremacy on the battlefield. A fighter jet or tank will be superior to adversary platforms in combat to the degree it can maintain an edge in these software-defined capabilities.
U.S. military leaders have begun to frame these critical concepts in their own terms of joint all-domain command & control or human-machine teaming. In pursuing them they should adopt a set of trends and best practices from the commercial industry to sidestep the technical hurdles of fielding these platforms with traditional, hardware-centric models. Onboard software is where we, or our adversaries, will hold the decisive edge. In order to realize the potential benefits of the digital revolution, the U.S. military too can embrace a software-defined vehicle design in future military platforms.
At Applied Intuition Defense, we build best-in-class digital engineering tools for leading autonomy programs - from the world’s largest automotive manufacturers (OEMs) to the Department of Defense. Contact us to learn more about how we can accelerate your team.
.webp)
.webp)
