Distributed Decision-Making for Real-Time In-Building Evacuation Guidance

The guidance in building evacuation during and after a disaster is an important and challenging problem, as the evacuees usually do not know the best route for evacuation and the static exit signals do not necessarily guide the evacuees to the right direction in dynamic situations. To find a best path in real-time within buildings is nontrivial. This paper proposes a distributed decision support system that efficiently computes the routes in real-time for building evacuation scenarios. Dijkstra's algorithm and Dynamic Programming are used for the route analysis. In the proposed approach, the physical system is composed of sensor nodes and decision nodes. The decision nodes, which are located at suitable locations for dynamic guidance, are equipped with information displays. The approach is evaluated in digital models and the authors built a 3D model with wall properties to be simulated with the Fire Dynamic Simulator (FDS). Fire spread and temperature distribution in the building under a mesh model are used as sensor input to test the proposed algorithm.