Emergency situations in public buildings: How to know where persons are to be rescued

In this paper an indoor detection system is described that is able to detect persons in emergency situations such as fires and is robust even if some sensors fail. Actual occupancy values are measured by data fusion and bias correction algorithms, and extended with credibility values of the sensor measurements. Based on this credibility value of sensor nodes, a metaheuristic estimates plausible "occupancy estimations" of the building. If sensor telemetry is invalid or implausible, credibility is degraded, otherwise raised, letting the system dynamically react to defect devices. The sensor network is designed openly, making it easy to integrate new, previously unknown, sensors to the mesh. To validate the result, disaster scenarios were tested in a simulated public building environment.