Artificial intelligence and simulation

Solutions for Current and Future Data Management Challenges

Data volumes increase both within complex onboard systems and through growing connectivity with external sensors, command structures, and other platforms. Particularly in military development and operational testing, these data volumes demand advanced analytics and fault diagnosis capabilities. “Modern defense platforms are highly complex due to their software-driven architecture and diverse assistance and autonomy functions. Therefore, it is necessary to efficiently store, manage, and analyze data—often in the petabyte range. To meet these challenges, we developed IAV Merida,” explains Dr. Remo Lachmann, data expert at IAV. This SaaS solution simplifies handling large data volumes during the development and operational lifecycle of military systems.

Increased efficiency and risk minimization through AI-based testing procedures

Efficiency, safety, and cost control play a central role in the development of defense technologies. In the defense sector in particular, real-world test drives with prototypes are often costly, risky, and require a lot of manpower. The challenge is to meet the highest quality and safety standards while staying within budget. The use of virtual test drivers, based on reinforcement learning and modern simulation environments, offers enormous potential here: development cycles can be shortened, real-world test drives minimized, and the costs for physical prototypes significantly reduced.

Virtual analysis of vehicle dynamics: A new era of development

Driving dynamics are the key factor for the maneuverability, safety, and comfort of vehicles—and therefore also of central importance in military applications. The ever-increasing performance requirements and growing complexity of modern vehicles and assistance systems are making traditional testing methods increasingly inefficient.

“By integrating AI-based test drivers into digital development, we can simulate entire scenarios under extreme conditions that would be either too risky or simply impossible for human drivers,” explains Simon Olma, project manager for Virtualize It at IAV.

Reinforcement learning as a virtual test driver

The new approach combines simulation technology with artificial intelligence: a reinforcement learning agent acts as a virtual driver and automatically explores the driving dynamics limits of a vehicle. Using a specially developed co-simulation environment, the agent interacts seamlessly with the simulation software and learns to specifically target critical driving conditions.

Focus on cost efficiency and safety

Minimizing costs and risks is essential, especially in the defense sector. Virtual tests replace expensive and dangerous test drives, improve reproducibility, and enable extreme scenarios to be played out that would be hardly justifiable in real-life tests. The number of physical prototypes can also be significantly reduced through targeted virtual analyses.

Technical implementation and innovative methods

At the heart of the system is a virtual test driver that uses a combination of AI and vehicle simulation to specifically simulate various driving situations. The AI agent takes over all important control commands such as acceleration, braking, and steering and can specifically target critical limit areas.

“The particular strength of our approach lies in the fact that the agent acts not only reactively but also proactively – this allows us to map the interaction of dynamic systems with a new level of precision,” says Johannes Dornheim, AI expert and team manager at IAV.

Performance is measured using special evaluation methods and shows that the virtual test driver achieves the level of human experts in many areas, enabling realistic, safe, and reproducible tests on the digital model.

Evaluation and outlook

Simulation results prove that RL-based test drivers are not only more efficient but also more versatile than human drivers. The method is scalable, adaptable to different vehicle systems, and opens up new possibilities for the development and validation of vehicle dynamics systems – a decisive factor in meeting the pressure for innovation in the defense sector at reasonable costs and risks.