TY - GEN
T1 - Evacuating routes in indoor-fire scenarios with selection of safe exits on known and unknown buildings using machine learning
AU - Agnihotri, Aakanksha
AU - Fathi-Kazerooni, Sina
AU - Kaymak, Yagiz
AU - Rojas-Cessa, Roberto
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/7/2
Y1 - 2018/7/2
N2 - We propose a scheme for calculation of indoor evacuation routes of a single-floor building in the event of a fire that uses selection of a potentially safe exit before evacuation starts. We refer to potentially safe exits to those that have high probability of being accessible at evacuation time. The scheme pre-calculates whether an exit may be reached by an occupant before being reached by the fire. Among those exits, the one with the shortest distance to the occupant is selected. This exit pre-selection improves evacuation success ratio by avoiding destination changes during evacuation. This approach is applicable to a building where the floor plan is known and analyzable. With such an evacuation routing scheme, we study the applicability of our approach to cases where the floor plan has not been characterized and yet we may be able to perform route calculation in a fast manner. To do so, we use machine learning with data obtained from a floor plan where the evacuation success ratio has been analyzed and apply it on the new floor plan. This approach indicates floor plan similarities and it is used to rapidly estimate evacuation routes with high probability of a successful evacuation. We show how floor similarity accuracy estimation increases with the use of data from an increasing number of analyzed floor plans.
AB - We propose a scheme for calculation of indoor evacuation routes of a single-floor building in the event of a fire that uses selection of a potentially safe exit before evacuation starts. We refer to potentially safe exits to those that have high probability of being accessible at evacuation time. The scheme pre-calculates whether an exit may be reached by an occupant before being reached by the fire. Among those exits, the one with the shortest distance to the occupant is selected. This exit pre-selection improves evacuation success ratio by avoiding destination changes during evacuation. This approach is applicable to a building where the floor plan is known and analyzable. With such an evacuation routing scheme, we study the applicability of our approach to cases where the floor plan has not been characterized and yet we may be able to perform route calculation in a fast manner. To do so, we use machine learning with data obtained from a floor plan where the evacuation success ratio has been analyzed and apply it on the new floor plan. This approach indicates floor plan similarities and it is used to rapidly estimate evacuation routes with high probability of a successful evacuation. We show how floor similarity accuracy estimation increases with the use of data from an increasing number of analyzed floor plans.
KW - emergency network
KW - emergency scenario
KW - evacuation
KW - fire
KW - sensor network
UR - http://www.scopus.com/inward/record.url?scp=85067097730&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85067097730&partnerID=8YFLogxK
U2 - 10.1109/SARNOF.2018.8720478
DO - 10.1109/SARNOF.2018.8720478
M3 - Conference contribution
AN - SCOPUS:85067097730
T3 - 2018 IEEE 39th Sarnoff Symposium, Sarnoff 2018
BT - 2018 IEEE 39th Sarnoff Symposium, Sarnoff 2018
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 39th IEEE Sarnoff Symposium, Sarnoff 2018
Y2 - 24 September 2018 through 25 September 2018
ER -