Am Montag, den 10. Februar 2025, um 16:15 Uhr hält
Rabeaeh Kiaghadi
Universität Oldenburg
im Rahmen ihrer beabsichtigten Dissertation einen Vortrag mit dem Titel
Disasters, Dynamics, and Decisions: Strategic planning in stochastic dynamic environments
Der Vortrag findet im OFFIS, Escherweg 2, Raum D21 statt.
Abstract:
In disaster scenarios, optimal scheduling of rescue operations poses a challenge, as perilous injuries incurred by victims induce an instationary planning problem with individually varied and dynamically changing survival probabilities. Autonomous search-and-rescue (SAR) strategies in such environments require a precise handling of both discrete events and continuously (possibly) degrading victim health conditions. In this thesis, a new model class, Interactive Markov Chains with Clocks (IMCC), has been proposed that merges the concurrency and stochastic behaviors of Interactive Markov Chains with the precise timing mechanisms of Timed Automata. Unlike previous models, IMCCs accommodate continuously varying, clock-dependent transition rates — a critical feature to capture the dynamic, time-inhomogeneous behavior of rescue scenarios.
Building on a rigorous analysis of decidability and computability properties of IMCCs, a heuristic strategy synthesis algorithm has been developed and implemented to systematically construct near-optimal real-time patient visit sequences in a variety of stochastic environments, where optimality refers to maximization of the expected number of rescued victims. More generally, the approach tackles the complexity of time-critical decision-making in continuously time-inhomogeneous environments. The proposed IMCC model and its accompanying algorithms provide rigorous development of autonomous strategies under uncertainty and tight time constraints.
Betreuer: Prof. Dr. Martin Fränzle