Abstract

We consider the problem of guiding evacuees in real-time during an emergency, focusing on pedestrian evacuation during a fire. We model the problem as a dynamic network flow with nodes and edges that expire over time. We compare an exact approach against a number of different heuristics for a greedy approach, and demonstrate that the heuristic framework can provide high quality plans in real time. We then develop an exact approach using an implicit representation of the time-expanded graph, and an anytime version, and show that for realistic real-time constraints, we can deliver solutions that are within 99.5% of the optimal solution. This is joint work with Tarik Hadzic and Cormac Sreenan.

Bio: Ken Brown is a Senior Lecturer in Computer Science and Deputy Director of 4C, the Cork Constraint Computation Centre. He is co-PI on CTVR , an Irish Telecommunications Research Centre. His research is in the area of constraint programming and applications of Artificial Intelligence, with a recent focus on applications to sensor networks and wireless communication networks