Fundamental Concepts and Evolution of Wi-Fi User Localization: An Overview Based on Different Case Studies

Abstract

Indoor positioning poses a number of challenges, especially in large and complex buildings. Several effects, such as signal attenuation, signal fluctuations, interference, and multipath play a decisive role in signal propagation. The severity of each challenge depends on the method and technology adopted to perform user localization. Wi-Fi is a popular method because of its ubiquity with already available public and private infrastructure in many environments and the ability for mobile clients, such as smartphones, to receive these signals. In this contribution, the fundamental concepts and basics and the evolution of Wi-Fi as the most widely used indoor positioning technology are reviewed and demonstrated using four different conducted case studies. Starting from an analysis of the properties of Wi-Fi signals and their propagation, suitable techniques are identified. The mathematical models of location fingerprinting and lateration are consolidated and assessed as well as new technology directions and developments highlighted. Results of the case studies demonstrate the capability of Wi-Fi for continuous user localization also in dynamic environments and kinematic mode where the user walks with a usual step speed. However, to achieve acceptable localization accuracy, calibration of the devices is required to mitigate the variance problems due to the device heterogeneity.

Keywords: combination and fusion of techniques; indoor smartphone user localization; lateration; location fingerprinting; positioning algorithms; received signal strength RSS; round trip time RTT measurements.

Figure 1

Free space path loss (FSPL) in dependence on the range for three different frequencies.

Figure 2

Signal strengths time series of long-term recordings with a smartphone to three Access Points (APs) for the 2.4 and 5 GHz frequency bands.

Figure 3

Influence of the human body on the orientation at two test locations CP 13 and CP 14.

Figure 4

Signal strengths time series of long-term recordings with six smartphones of one AP.

Figure 5

Multivariate linear regression for offset determination of individual smartphones in the system calibration of heterogeneous devices for (a) the 2.4 GHz and (b) 5 GHz frequency band.

Figure 6

RSS values before (a) and after (b) the calibration.

Figure 7

Positioning approach for the example of four visible APs with (a) measurements from the off-line training phase and (b) current measurements in the on-line positioning phase.

Figure 8

Radio maps of RSS distributions in the entrance area of a multi-story office building with (a) arithmetic mean, (b) median, (c) minimum, and (d) maximum RSS of four orientations. For the location of the six Raspberry Pi units (RPs) and the test locations, see Figure 14a.

Figure 9

Example of a radio maps stack in the datacube for four sensed APs.

Figure 10

Allocation of positioning scans to the stored RSS scans in the training fingerprinting database DB [42].

Figure 11

Resulting mean deviations from the ground truth of the K-nearest neighbor (KNN) approach for different K values ranging from 1 to 15. The measurements were carried out along two different trajectories located on the ground floor of a multi-story office building (see Figure 15 for their location).

Figure 12

RSS to range relationship of the mean of all four orientations (a) for four smartphones (ID1, ID2, ID3, and ID7) and (b) in all four orientations for smartphones ID7 on the right.

Figure 13

Differential Wi-Fi (DWi-Fi) operational principle with four APs.

Figure 14

Indoor trajectory (a) and deviations from the ground truth without and with calibration for seven different smartphones (SPs) on two waypoints CP 31 (b) and CP 32 (c)

Figure 15

Two trajectories (a) in the tests site on the ground floor of a multi-story office building leading from outdoors through the main entrance (b) to the foyer (c), the classroom VII (d), and an area with desktop computers (e) to the inner courtyard.

Figure 16

Resulting trajectories of a user walking along with usual step speed (two results for trajectory 1 (a,b) and two results for trajectory 2 (c,d)) with long scan duration of the smartphone (a,c) and short scan duration (b,d) resulting in significantly different numbers of trajectory waypoints.

Figure 16

Resulting trajectories of a user walking along with usual step speed (two results for trajectory 1 (a,b) and two results for trajectory 2 (c,d)) with long scan duration of the smartphone (a,c) and short scan duration (b,d) resulting in significantly different numbers of trajectory waypoints.

Figure 17

Comparison of the maximum deviations of the position estimates for kinematic positioning using the one-slope, multi-wall model, and differential approach (DWi-Fi) for three different smartphones (SP) in the different areas of the test site. The trajectory leads from the foyer through a classroom VII and an area with desktop computers (red trajectory in Figure 15a).

Figure 18

Comparison of the average deviations of the Wi-Fi RTT measurements on 12 test points evenly distributed in a grid with a spacing of 6 m between them.