Artificial intelligence enables innovative digital solutions to support decision-making – for the benefit of all stakeholders. HENSOLDT is also researching these intelligent defence applications of the future in close cooperation with specialized innovation partners.
The use of artificial intelligence (AI) is one of the most important key technologies of our time. Today, it already analyzes visual data and, with the help of sensors, can smell, feel, and – in the form of Alexa, Siri, and the like – even hear and speak. It is being used in a variety of ways and has already become an integral part of areas such as production, finance, and health care as well as our daily lives.
By its very nature, application in defence systems places particularly complex demands on AI. HENSOLDT uses AI in nearly all of the company’s core technologies and continues to expand its AI expertise. Only with AI can the ever-increasing volume of data be processed and evaluated quickly and reliably, often in milliseconds. This is the only way to achieve a level of automation that enables timely, impact-optimized action while conserving resources and minimizing collateral effects. Humans are relieved by the preparation of options for action, but at the same time should and must always remain involved in the control. AI should support, but must never decide, independently. That is why HENSOLDT is committed to the principles formulated by the European Commission regarding trustworthy AI systems.
In order to explore the future role of AI in defence applications, HENSOLDT is already taking the next step. In cooperation with the Helmut Schmidt University of the Bundeswehr in Hamburg, possible deployment scenarios for the AI of the future are being developed and tested. At the center of this: a form of AI that can take a predictive perspective for the first time. The research project is called „GhostPlay“ and is funded by Zentrum für Digitalisierungs- und Technologieforschung der Bundeswehr (DTEC.Bw).
The goal of “GhostPlay” is no longer solely to evaluate current data alone more effectively and quickly. Instead, AI is to analyze the potential behavior of potential adversaries, learn from it, and develop and recommend possible countermeasures in advance. AI thus enables tactical behavior under highly complex conditions, in which, analogous to a chess player, several of the opponent’s moves are anticipated and used as the basis for one’s own decision-making. This opens up completely new dimensions: In air combat, for example, AI could anticipate the upcoming flight maneuvers of another aircraft. The navy could better protect its ships from unmanned drone swarms and target its torpedoes more effectively. On the ground, sensor-effector networks in air defence systems, for example, could be widely controlled in their interaction to respond predictively to potential attacks. Even from space, satellite-based sensor technology is to be effectively integrated into all decision-making processes.
However, not every kind of AI can automatically be used for every operational scenario. That is why “GhostPlay” uses various AI approaches, from which the relevant ones are then combined. In ground operations, for example, the adversaries – and thus their AI – will have to learn from each other in the future, and thus the different AI will also have to incorporate a learning dynamic. Drone swarms, on the other hand, still behave according to relatively simple rules today. The challenge here lies more in the large number of agents taking part, each of whom must be reliably recognized. In the future, however – and this is also part of “GhostPlay” – swarms of drones will also be controlled by AI, for whose defence, in turn, special algorithms must be developed.
In this context, HENSOLDT builds on its profound experience with the application of AI in a wide variety of fields: in the networking of sensors and effectors. When evaluating reconnaissance data from a wide variety of sensors, such as the combination of passive radar, mobile radar, and “
SPEXER 360-degree radar in air defence or sensor fusion in the See Through Armour System. With the use and processing of data, also from additional sources tailored to the concepts of operations of its customers. In addition, HENSOLDT is also contributing its expertise in simulation techniques in product development and resource management to the project.
The “GhostPlay” team led by HENSOLDT also includes: “21strategies”, a company specializing in AI-supported, predictive decision-making, and Borchert Consulting & Research AG for strategy, concept, and scenario development.
Together, the team is developing AI systems that will be trained in simulations as part of the study project. All assets involved in the simulation are thereby empowered by AI. They will then compete against each other and learn independently in a military context in order to make future-oriented decisions. The result is a clash of tactical, intelligent systems that learn from each other: A recognizes the behavior of B, whereupon B in turn challenges A with a new tactic. In this way, future military operations will be analyzed by AI, new tactical approaches will be derived, the application of new technologies and systems will be evaluated, and the dangers to humans and machines will be minimized while taking ethical principles into account. At the same time, research and documentation will be conducted on the potential risks that may arise.
The (Perfect) Third Wave HENSOLDT has been using AI algorithms in its products for years. So far, the focus has been on the first two waves of artificial intelligence: machine learning and deep learning.
Machine learning uses typical patterns to enable computers to classify and assign new data based on these patterns.
Deep learning goes one step further. Here, the algorithms are much more complicated and usually inspired by human neural networks. This means that far larger amounts of data can be analyzed.
Projects like “GhostPlay” are now opening the door to the third wave of AI: reinforcement learning, which goes beyond fact-based decision situations. In this process, an AI is trained to recognize and evaluate very complex decision structures and make the best possible decision based on them. To do this, it not only analyzes the data, but also puts it into context and constantly learns.
For defence applications, reinforcement learning can be used in the future primarily in very complex situations to enable even more intelligent decisions independent of domains.