AI-powered decision support system for wildfire disaster response and management (Disasteraware)

Date

2026-8-31

Author

Akman, Çağlar

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

23
views

13
downloads

Cite This

Wildfire decision support systems use predefined decisions, cannot adjust the rule bases during an incident, and rarely expose the reasoning. DisasterAware closes these gaps with generative concept-based fuzzy reasoning. One local agent per region maps confidence-gated observations onto decision concepts and, through a Takagi-Sugeno rule base, onto interventions; a coordination layer ranks regions and shares capacity. The main contribution is an open decision space. A staged loop tunes rules, writes new ones for uncovered situations, and, when the base has no answer, runs a generative model that may propose rules, blend interventions, and compose new ones from verified physical effects. Each modification must pass a forecast test, and a generated one must also clear deterministic gates so every decision stays traceable. The simulation core reproduces real fire propagation with 86 to 94 percent agreement. Over 1000 simulations, DisasterAware extinguishes 52 percent of the fires within the episode, from a static rule base extinguishes 10 percent. It leaves 28 percent fewer people affected. Against no intervention, it cuts the burned area by 70 percent. Producing a decision takes less than 0.2 percent of the decision cycle.

Subject Keywords

Decision Support System, Concept-Driven Fuzzy Inference,, Open Decision Space, Generative Rule Proposal, Traceability

URI

https://hdl.handle.net/11511/121036

Collections

Graduate School of Natural and Applied Sciences, Thesis