The functioning of modern societies relies on a largely disruption-free global logistics and services network, which can be severely destabilised by extreme weather events, geopolitical conflicts, or local production concentrations. A continuous forecast providing sufficient lead time to take countermeasures has so far been hindered by the complexity of goods and service chains, as well as the challenge of incorporating current information — e.g. from news sources. PROVIDER addresses precisely this gap by acting proactively, thus enabling preventive measures in risk and crisis management — in direct alignment with the funding policy objectives of the BMFTR for strengthening a resilient society.

The PROVIDER project pursues the central goal of developing an innovative methodology and system implementation for the continuous, dynamic simulation enabling early prediction of supply shortages. Unlike conventional post-hoc analyses or static scenarios, the aim is to create a permanently operating system that monitors the supply security status for all known commodity groups and generates reliable forecasts with a time horizon of several months. The system explicitly targets critical goods and services outside classical KRITIS sectors that have so far been insufficiently considered in crisis planning.

PROVIDER integrates four closely interlinked key components: First, the dynamic generation of extensive Knowledge Graphs from heterogeneous data sources (including DATEV data and public datasets) that structurally capture global goods and services networks. Second, self-learning agents are trained using Deep Reinforcement Learning and instantiated in an agent-based simulation environment that realistically models production, trade, transport, and services. Third, Large Language Models (LLMs) enable intelligent parameterisation of the simulation by analysing current news reports, extracting relevant risk factors, and identifying potential bottlenecks at an early stage. Fourth, the Adversarial Resilience Learning (ARL) methodology ensures manageable and efficient computation of highly complex, cross-sector value chains and enables targeted resilience analysi