A Technique for Finding Indoor Rescuer using GNSS and Barometer Sensors on Android Smartphones in Building Disaster Situations

Abstract

Listen

Translate article

In disaster situations such as fire or building collapse, the exact location of the victim must be identified so that rescuers can quickly rescue. There are methods based on audio and video to determine the location of the victims, but this method is not effective for hot flames and thick fog. Another location tracking method is to utilize the Global Navigation Satellite System (GNSS) mounted on the victim’s smartphone. The GNSS satellites have the advantage of being able to provide real-time location information to the entire earth without being affected by time and weather, so they are highly utilized for rescue activities. In general, in the case of a positioning method using GNSS satellites, a minimum number of satellites of 4 or more is required, with 3 satellites for user location determination and 1 satellite for user clock bias error. In the event of a fire inside the building, victims request rescue and wait for rescue around the window. In the case, the GNSS satellite signals are blocked by walls inside and outside the building, so only one direction signals are received, which is the main reason for increasing the satellite Dilution of Precision (DOP) values. The increase of the DOP value directly affects the user positioning results, and the positioning result spreads widely toward the satellites. In addition, the visible environment in one direction often does not meet the minimum number of satellites for positioning. Fig. 1 shows the number of GNSS satellites observed with a smartphone in an indoor window environment. The change in the number of GNSS satellites over the entire interval is very large and several intervals observed under the minimum positioning condition (# of satellites < 4) are observed. Even at about 400s, only one is observed from GPS satellites. Therefore, in this case, positioning using only GNSS satellites is not possible and additional measurements are required. Since Google announced that it offers raw GNSS measurements on Android 7.0 and higher operating systems, research on improving location performance in smartphones has become possible using measurements such as pseudorange, carrier, doppler, and Signal-to-Noise Ratio (SNR). Since in previous versions only calculated location information was available, location information obtained from smartphones had errors of tens of meters, and positioning was not possible in less than four poor GNSS visible satellite environments. Since the Google announcement, it has been possible to conduct research on integrating positioning by combining raw GNSS measurements from smartphones with various motion sensor measurements. Smartphones are equipped with various motion sensors such as acceleration sensor, gyro sensor, geomagnetic sensor, and barometric pressure sensor. Among them, the barometer sensor is a sensor that detects air pressure and is mainly used to measure the altitude by detecting the amount of compressed air according to the sea level. In this research, to compensate for the lack of GNSS visible satellites, we used a method using altitude measurements from the barometer sensor built into Android smartphones. In the case of barometer altitude measurements, since it is a Latitude-Longitude-Altitude (LLA) coordinate system, it must be changed according to the Earth-Centered Earth-Fixed (ECEF) coordinate system in order to add it to the GNSS observation matrix. The observation equation with the added barometer altitude measurement enables positioning even in situations where the minimum number of satellites is not satisfied. To prove this algorithm, we performed a simulation using satellite data in one direction (mask azimuth angle = 110 ~ 250 degrees) and compared the results. Fig. 2 below is a graph comparing DGNSS and DGNSS+DBaro positioning results. In the case of the DGNSS positioning results, only the satellites observed from the south were used, it can be confirmed that they are widely spread to the north and south. The RMS error of DGNSS was found to be a very large value, such as 18.3m horizontally and 29.9m vertically. In the case of the DGNSS+DBaro positioning result, the vertical direction error was greatly improved by the addition of the barometer's altitude measurement. This has the effect of dispersing the GNSS measurements used in the vertical direction in the horizontal direction, and has a great influence on the improvement of the horizontal direction error. In the case of the DGNSS+DBaro positioning results, the RMS error compared to DGNSS was 30.9% horizontally and 1.8% vertically.