BMBF-Projekt Risikogesteuerte Umfeldexploration für Sicherheitsaufgaben
Prof. Dr. Nicole Bäuerle (Koordination)
Karlsruher Institut für Technologie
Institut für Stochastik
D 76131 Karlsruhe
Dr. Jürgen Geisler
Abt. Interaktive Analyse und Diagnose
Fraunhofer-Institut für Informations- und Datenverarbeitung (IITB)
D 76131 Karlsruhe
Dr. Peter Klausmann
Prof. Dr. Norbert Link
D 76133 Karlsruhe
Coping with threats such as terrorism, crime, natural hazards or industrial accidents requires early detection and correct assessment of critical situations, followed by appropriate actions to avoid potential damage to person or property. We here focus on threat prevention in closed properties of traffic infrastructures (railway stations, harbours, airports) or logistics centres. To early detect and prevent damage, the property is continuously monitored by
sensors and security staff. We propose to model the threat situation of the property as a continuous-time Markov decision process (CMDP). The CMDP-model is used as core of a decision support system (DSS) which assists the security staff in two aspects: it facilitates to identify the threat situation of the property and suggests the cost-optimal action by which the property will be kept in or returned into a safe state.
A decision support system should help to identify the current threat situation and suggest how to best deal with the current threats. For that purpose, the decision support system requires a description of the property's threat situation. As a solution, a CMDP-model for surveillance applications for closed properties is introduced. It is based upon a concept of threat events, which can occur randomly in the sectors of the property. The model defines a risk assessment for the property as well as admissible actions to deal with the threat. From the model, the optimal policy, the core of the decision support system, can be derived. A numerical example for an airport operation has been computed so far. Since the expected discounted cost-optimal policy has no simple structure, it could hardly be predicted by the decision maker himself. Thus, it is beneficial to base the decision support system upon the introduced model.