IntelPlaMo

Ports, railway lines, airports and other critical infrastructure must operate reliably around the clock. If damage to tracks, fences or runways is detected too late, it can lead to accidents, operational disruptions or even deliberate security incidents. Current monitoring systems rely solely on fixed sensors and offer insufficient protection against cyberattacks on the AI components used. Furthermore, there is a lack of solutions that effectively integrate human operators into automated decision-making processes.

Drones and AI: a reliable surveillance duo

As part of the IntelPlaMo project (Intelligent condition analysis for mission planning and monitoring of safety-critical infrastructure), the German Aerospace Center (DLR) is developing innovative methods for the automated surveillance of critical infrastructure. This involves combining data from ground-based sensors with data from mobile autonomous systems, particularly Unmanned Aerial Vehicles (UAVs). This multimodal sensor data is then fed into a back-end data ecosystem where it is processed using AI-based algorithms and visualised in a digital control centre.
A key feature of this approach is that AI systems are secured against potential attacks from the outset. Adversarial attacks, which involve the targeted manipulation of AI decisions, are systematically analysed and countermeasures are developed. The reliability of the results is presented transparently. This allows human operators to assess the reliability of an AI's assessment at any time.

From simulation to practice

The developed methods are being tested in two real-world application scenarios. First, the procedures are being tested and validated using a simulation based on the digital twin of the Braunschweig Port railway. This will be followed by a live demonstration at the National Test Centre for Unmanned Aerial Systems at Magdeburg-Cochstedt Airport in summer 2028.

Contribution Institute for AI Safety and Security

The Institute is responsible for safeguarding AI systems against attacks and malfunctions. Its work focuses on continuously monitoring AI decision-making chains, developing software artefacts for attack detection and creating a structured security case for using AI in safety-critical applications.