[Back]

G. Retscher, E. Mok, T. Hecht:
"Smartphone Altitude Determination Using Embedded Barometric Pressure Sensors";
Talk: 10th International Symposium on Location-Based Services LBS 2013, Shanghai, China (invited); 2013-11-21 - 2013-11-22; in: "10th International Symposium on Location-Based Services LBS 2013", (2013), 3 pages.

Modern smartphones have become very popular for users also in the field of Location-based Services (LBS) and for different navigation tasks. Thereby location determination can be performed in various ways. In most cases the location determination relies mainly on GNSS. If GNSS is inaccurate or fails, the embedded motion sensors in the smartphones such as three-axis MEMS (short for Micro-Electro-Mechanical Systems) accelerometers, gyroscope and digital compass or magnetometer can overtake the task of localization. The motion sensors are usually employed for dead reckoning (DR). In this case, continuous recording of the sensor data is employed to estimate the current location of the user. From a given start position (e.g. determined from GNSS) the distance travelled is derived from accelerometer measurements (e.g. the accelerometer is used to count the steps of a pedestrian user or to derive the distance travelled of a vehicle) and the heading from the measurements of the magnetometer and eventually a gyroscope. Several tests have been performed by the authors to investigate the performance of five different smartphones for location determination using GNSS and the motion sensors in combined indoor/outdoor environments. Thereby the selected test environments included an outdoor reference trajectory in Vienna and a bus route in urban canyons in Hong Kong where very often satellite signals are obstructed by tall buildings or large structures causing insufficient number of GNSS measurements for successful position determination. In addition, indoor tests were performed in an office building where the user started at a known positions outside the building (e.g. obtained from GNSS observations) and moved indoors to various rooms on the same floor. In all tests it could be proven that the combination of the embedded motion sensors in smartphones can improve the positioning performance and provide continuous position determination in 2-D in the case if GNSS positioning fails for a short time period. An important requirement, however, for positioning in indoor as well as outdoor environment with the motion sensors is a proper calibration of the sensors. Apps have been developed to calibrate the step length of the user and setting of thresholds for acceleration measurements which influence how sensitive the application is in step detection. Moreover, correction for the gravity effect on the x-, y- and z-axes of the smartphone´s local coordinate system is the key to correct the determination of accelerometer-derived distance travelled. In addition, the compass has to apply magnetic to grid north correction to improve the heading determination. For indoor environments also the use of Wi-Fi location determination with fingerprinting has been investigated as a positioning method which can provide absolute coordinates and therefore is able to correct the drift rate of the motion sensors used for dead reckoning. To summarize the test results it can be said that the overall performance and positioning accuracy in combined indoor/outdoor environments can yield several metres in positioning accuracy of a mobile pedestrian user in 2-D.
Only little attention, however, has been drawn so far to the fact that also the determination of the altitude of the user can be of very high importance, for instance, if a user should be located on the correct floor in a multi-storey building. Recently also atmospheric (or barometric) pressure sensors can increasingly be found in modern smartphones. The investigation of their capability and performance for height estimation of a user is the main aim of the study described in this paper. Two different tasks are investigated: Firstly the use of the sensors in outdoor environment if GNSS is only capable to provide a 2-D positioning solution or for continuous positioning using dead reckoning if GNSS fails. The second application scenario is indoors where no height determination is available from GNSS. In a previously conducted study at the Vienna University of Technology for the development of a pedestrian navigation system which works in combined indoor/outdoor environments the augmentation of the system with a barometric pressure sensor has been investigated. Altitude determination in the so-called NAVIO system is performed relatively with a barometric pressure sensor from a given start height (e.g. obtained from GNSS outside the building or from a known height point in the indoor environment). As the user walks inside the building and up the stairs or elevator to other floors the altitude is determined relatively to the given start value. From the difference and changes of the air pressure in the sensor altitude differences can be calculated using a simple relationship between the pressure changes and height differences. This process involves pressure observations at the two stations which geographic location is approximately known. For the conversion of the air pressure in a height difference also the mean value of the temperature at both stations is required. The investigations and tests of the barometric pressure sensor integrated in the NAVIO system in a multi-storey office building showed a satisfactory performance and maximum deviation from the truth for the altitude determination of around 1.0 m for over 90 % of the observations. Therefore the employed barometric pressure sensor in the NAVIO system was always capable to determine the correct floor in which the user was currently located.
In this study the low-cost barometric pressure sensors in modern smartphones are tested in a similar manner to investigate if they are also capable to provide such performance and accuracy for altitude determination. As most smartphones also incorporate low-cost temperature sensors they can also provide mean temperature differences between the stations required for the conversion of the air pressure changes into altitude differences. In addition, the sensor drift rates and the required calibration procedures are investigated and discussed. One approach to test the drift rates of the barometric pressure sensors is long-term observation on benchmarks in a laboratory environment as well as outdoors. For that purpose sensor test measurements are conducted over a duration of up to a few hours in static mode. In the following it is investigated if a functional connection for the relationship between the observed air pressure changes and the altitude differences can be derived. Such a relationship can be described, for instance, using characteristic curves. If such a curve exists and the functional connection is also linear in a certain range then the pressure differences can be converted into height differences with the required precision. It must be noted that the altitude deviations depend also on the time of day, e.g. higher deviations occur during noon where usually more people are inside the building and larger variations of the air pressure are obtained caused by higher air circulation due to frequent opening of doors and windows. For that reason observations in a building are carried out during different times of the day and different absolute air pressures. The tests can be performed along a 3-D reference trajectory in a multi-storey office building of the Vienna University of Technology. This trajectory has been determined by conventional surveying techniques and incorporates 26 reference points on different building levels. Measurements can be started on a reference point on the ground floor and go up onto the roof of the building where several survey benchmarks are available or also vice versa. From the obtained functional relationship a correction for the pressure observations in the building might be derived to improve the precision of the altitude determination. Then it is investigated if the height determination is improved by applying these correction terms to the observed pressure changes. Apps will be developed for the calibration and correction of the sensor observations and calculation of the integrated 3-D positioning solution. Several test runs in the building where the user walks to different floors and changes between floors levels are conducted and analyzed. Their results are presented in this contribution. The investigations will lead to reliable 3-D ubiquitous positioning algorithms for location determination using modern smartphones for LBS and other navigation applications. Due to further development of new advanced low-cost sensors we believe that the use of multi-sensor solutions including a barometric pressure sensor which provide 3-D location capabilities in outdoor as well as indoor environments are a necessary requirement for modern smartphone navigation services.

Smartphone Location, Motion Sensors, Barometric Pressure Sensor, Altitude Determination, Indoor/Outdoor Positioning