Wi-Fi Location Fingerprint Use in Indoor Positioning

UESTC Glasgow College 2017 Yan Heyun

Abstract:

For GPS and other satellite navigation systems, it is difficult to locate positions of indoor objects. Simultaneously, Wi-Fi usually exists in most indoor environments, so a cost-effective method come up with—use Wi-Fi itself to give the position services without other devices. However, as Wi-Fi is not specifically designed for positioning, so the traditional time-based and angle-based location method is not feasible to Wi-Fi. Currently, location fingerprint has been widely researched and utilized in indoor situations. In this paper, Wi-Fi location fingerprint is briefly reviewed, situation of development of this technology and its social effect.

Key word:

Wi-Fi, positioning technology

Introduction:

The positioning of indoor environment is relatedly complex project because there are many problems needing to be solved. Because of severe signal attenuation, normal outdoor navigation and positioning facilities cannot work effectively and effectively indoors. Precision of positioning is also an unignorable problem. GPS may be able to point out which building the mobile device is in, but to be specific, users would like to get more details about locations of the targets, and this process requires much more precise map information and higher positioning accuracy.
Certainly, we can build a complete system to provide positioning indoors and increasing the positioning accuracy, but this requires a lot of costs, containing all sorts of devices. Therefore, if we want to use existed system sufficiently, we need the technology about location method based on wireless signal, for example, Wi-Fi. Wi-Fi is widely used in homes, hotels, airports, shopping malls and other buildings, which makes Wi-Fi become one of the most economic and realistic wireless technology utilizing in the field of positioning. However, Wi-Fi signal is not designed for positioning, is usually only one antenna, and the bandwidth is small. Furthermore, indoor complex environment of signal propagation makes the traditional based on measuring the difference of arrival time hard to achieve. If people try to handle it through installing directional antenna in the Wi-Fi network, it needs additional expenditure. Therefore, the location fingerprint method has already become the main method to solve above problems.

1. Positioning algorithm based on location fingerprint:

Location-based fingerprint is usually divided two types. One is a certain algorithm which compares signal characteristics (such as vector r) with pre-computed statistics in the database of plenty of fingerprints. The other is an algorithm of probability which calculates the probability that a signal feature belongs to a distribution (stored in a fingerprint database). Here are some basic methods, but they are not completed.

1.1 Certain localization algorithm:

The research of this algorithm was first done by Microsoft the fingerprint on Wi-Fi location in 2000. They use the value of the geometric distance between the RSS vector r and fingerprinting vector ρ to locate mobile devices. Assume that the location fingerprint is N-dimension, that is, there are N number visible APs, M grid points, so there are M fingerprints in the fingerprint database. And r and ρ distance definition for:
D=√(∑_(i=1)^N▒〖|r_i-ρ_i |〗)

In this way, the simplest locating algorithm can be described as follows:
In the M fingerprints in the fingerprint database, the closest fingerprint in the signal space to the observed value of RSS can be found, and the corresponding position coordinate can be used as the position of mobile device. The nearest neighbor in signal space could be found by this method, because the goal is to find a fingerprint in the signal space nearest the RSS observation value. Following figure shows the basic principle of this method, where the pentagram represents the RSS observation value and the circle represents the position fingerprint in the signal space. The decision boundary can be drawn using the Tyson polygon method, and the region contained in each Tyson polygon in the signal space is closest to the fingerprint at this position. The position of the fingerprint in the Tyson polygon region where the RSS observation value is located is used as the location result.

1.2 Probabilistic localization algorithm:

The earliest probabilistic localization algorithm based on Wi-Fi location fingerprint was proposed by Youssef et al. (2003). The basic idea was that simply using the statistics of an RSS sample (such as the mean value of RSS) might bring errors, as the actual RSS value should be a distribution. Therefore, we can use the joint probability distribution (there are multiple AP, so the joint probability distribution) as the fingerprint. Obtaining the joint probability distribution by collecting RSS samples is not an easy task because the correlation between RSS from each AP is not obvious. They assume that this is independent (which is reasonable), and then simply use the product of RSS’s marginal distributions as the joint distribution. Assuming that the observed RSS vector is r=[r1,r2,r3,…,rN], a grid point will be selected for the estimated position, and the maximum probability that this rr will be generated on the grid point will be. For a given r, bayes theorem can be used to estimate the location of mobile devices. The probability of r generation on a grid point can be calculated as follows:
P (Gridpoint | r) = [P (r | Gridpoint) ⋅ P (Gridpoint)]/ P ®

Calculate the probability of all grid points, and then select the maximum probability of the grid point as the location of the mobile device.

One problem that is not taken into account is that the same set of AP’s can not always be detected at all grid points. Youssef et al. (2003) first described this problem, Swangmuang and Krishnamurthy (2008b) considered different clustering methods. Youssef et al. (2003) 's work grouped grid points based on the equality of each AP. Grid points sharing the same set of AP are considered as a cluster, and the determination of the cluster is based on the probability that the AP can be seen at each grid point. Therefore, this method is also called “joint clustering” or JC technology as described by the author. By clustering fingerprints in the signal space, Swangmuang and Krishnamurthy could reduce the complexity of fingerprint search. Therefore, he hypothesised that the same set of AP could be seen at all locations.

By the way, Many other pattern matching algorithms can be applied to WiFi location fingerprinting. It includes Bayes Theorem, statistical learning theory, support vector machine, neural network and so on.

2. The social effect

This kind of technology could be widely used in service industry such as postal industry. As it can be used to located the precise position of users with terminals, it will make our daily life more convenient. People will not be confused about finding the exact positions in some large commerical centers. Although it might cause severe threat to safety of private information because a lot of devices support Wi-Fi, and Wi-Fi signal could be easily monitored, and if a network monitoring these devices is created for malicious purposes, a large number of devices can be tracked based on their MAC address and RSS value, which will lead to serious user privacy problems of personal information leakage, with the time going by, this problem will be solved in the future, caused by the further development of related technology.

Suggestion

Wi-Fi’s almost ubiquitous availability makes it an attractive location method (no additional hardware cost). Wi-Fi signals cannot use method of navigation based on time and angle, which makes location fingerprint a major choice for location. However, location fingerprinting requires complex data collection work and may need to be updated frequently with changes in the environment. In addition, due to the complexity and variability of radio transmission, the collection of location fingerprints is not an easy problem in itself. Some measurements, analyses and simulations have shown that some empirical methods can be used to reduce the workload of fingerprint collection. If the requirements of location accuracy are not high, other methods such as sub-region location or organic construction of location fingerprint can be used to reduce the workload of fingerprint collection.

Although location fingerprinting using only Wi-Fi has been proved to be a feasible solution for indoor accurate positioning, with the emergence of new technologies and the popularity of smart phones with nearby sensing capabilities, the final solution may be to use multiple positioning technologies for fusion. In the next few years, near-field communications, accelerometers in smartphones, and possibly bluetooth signals may be used as additional parameters to improve positioning accuracy while maintaining reasonable costs.

Reference

[1]: Ahonen, S., and P. Eskelinen. 2003. Mobile terminal location for UMTS. IEEE Aerospace and Electronic Systems Magazine, vol.18, no. 2, pp. 23–27.
[2]: Aksu, A., and P. Krishnamurthy. 2010. Sub-area localization: A simple calibration free approach. ACM MSWiM’10.
Bahl, P., and V. N. Padmanabhan. 2000. Radar: An in-building RF based user location and tracking system. IEEE INFOCOM 2000, pp. 775–784.
[3]: Husted, N., and S. Myers. 2010. Mobile location tracking in metro areas: Malnets and others. ACM CCS.