Armaments
Speeding Up Defense: The Software-Defined Revolution

As defense organizations continue to develop their systems in response to global political developments, they are faced with a key practical challenge: How to quickly deploy and update state-of-the-art software-based systems while also thoroughly meeting their safety requirements? On the one hand, modern military systems are urgently needed to support the safeguarding of national and international interests. On the other hand, stringent safety guidelines must always be maintained both during deployment and throughout operation. To add to this already-complex landscape, many of the desired features envision the use of artificial intelligence techniques, which are notoriously difficult to provide safety guarantees for. Clearly, a new approach is required to strike a balance between these factors and fulfill the most critical objective: To never endanger human life due to a lack of safety or reliability.

Software-Defined Defense

To speed up the deployment of new features in defense systems, the industry is increasingly turning towards the concept of software-defined defense (SDD). This consists of leveraging software solutions to make defense systems more flexible, thus moving away from older fixed-hardware approaches. In other words, the idea is to decouple the software from the hardware in defense systems, which allows functionality to be updated without having to modify any physical components.

To support this, a modular software architecture is required, with clearly defined interfaces between each of these modules. The use of this architecture allows for new functionality to be more easily integrated into the existing system, aside from increasing system adaptability and standardization. Moreover, it is particularly important for this architecture to support interoperability, as defense systems must often exchange information and coordinate actions when cooperating to accomplish more complex goals.

Three-Pronged SDD Implementation

The Fraunhofer Institute for Cognitive Systems IKS’ focus on combining safety approaches with the deployment of AI makes it uniquely positioned to support concrete implementations of SDD. It tackles the topic from three complementary directions:

• Adaptive architecture: The basis for the implementation of SDD is a modular architecture that supports the dynamic adaptation of the system. This includes a loop that continuously monitors the system’s state, analyzes any potential issues that may manifest, plans for behavior modifications as needed, and executes these modifications. Additionally, a central knowledge base provides the required information for carrying out these steps. This degree of adaptation supports SDD in that it allows system mechanisms to be updated without hardware changes or system downtime.
• Human-in-the-loop: The vision for SDD is not to replace the human decision-making process, but instead to provide the basis for collaborative decision that take both the perspective of users and the information from the system into account. This is crucial for ensuring that any adaptations to the system are ultimately decided on by the users, who are provided information from the system regarding the concept in which it operates.
• AI techniques: The use of AI also brings many advantages to the defense domain, allowing for more complex system analyses and behaviors. This includes, for instance the coordination of drones or other unmanned aerial vehicles (UAVs), as well as the implementation of learning capabilities for reconnaissance robots or other unmanned ground vehicles (UGVs). As such, the leveraging of AI techniques should be a priority in the implementation of SDD. In particular, the integration of AI within the adaptive architecture and its use with a human-in-the-loop mechanism to validate results can maximize the efficacy of overall system.

Where Safety Comes In

Of course, the implementation directions outlined above must abide by the necessary safety standards. This is accomplished by integrating safety-by-design principles (i.e., addressing intrinsic hazards as early as possible) throughout the entire SDD lifecycle, ensuring that adaptive behaviors and AI-driven decisions operate within predefined safety boundaries. Additionally, AI techniques are subjected to comprehensive validation and verification processes, including formal methods and safety case development. By employing these measures, it is possible to follow a SDD approach without compromising on the safety critical aspects of the system.

Next Steps

For organizations that are now contemplating the adoption of a SDD paradigm, starting this journey may feel overwhelming, as there are so many factors to simultaneously consider. Thus, a systematic approach must be adopted. First, a comprehensive assessment of the current architecture must be carried out to identify which elements present tighter coupling between software and hardware, as well as to understand the extent of the required modifications to the existing system structure. Then, the safety standards to be observed by the system must be mapped to the identified requirements to ensure that they are considered during the SDD lifecycle. Finally, a pilot implementation of the new SDD-enabled architecture can begin. By following this stepwise approach, organizations can strike a successful balance between fast innovation and safe execution.