Title: LANDMARC: indoor location sensing using active RFID
From: Percom’03
Note number: 200322
//
1 Abstract
LANDMARC, an indoor location sensing prototype system that uses RFID technology is proposed. The major advantage is that it improves the overall accuracy by utilizing the concept of reference tags. The experimental analysis has proved the feasibility of the approach.
2 Introduction
Three principal techniques for automatic location-sensing are introduced as triangulation, scene analysis and proximity. The approach proposed here belongs to the proximity one. The objective of this research is to develop a prototype indoor location-sensing system using easily accessible wireless devices. Then an overview to the subsequent sections is given in this part. Section 2 analyses strengths as well as limitations for several types of location-sensing systems, includes Infrared, 802.11, ultrasonic, and RFID. Thus, RFID technology is led out and then an overview of it is given in section 3.
As this article focuses on the LANDMARC approach, section 2,3 skip here.
3 LANDMARC Approach
Before the subsections in this part, scene design and devices of LANDMARC system are firstly explained.
In order to increase accuracy without placing more readers, the system employs extra fixed location reference tags to help location calibration, which serve as reference points.
There are 3 major advantages of LANDMARC:
1)There is no need for a large number of expensive RFID readers;
2)The environmental dynamics can easily be accommodated. In other words, the LANDMARC helps offset many environmental factors that contribute to the variations in detected range because the reference tags are subject to the same effect in the environment as the tags to be located.
3)The location information is more accurate and reliable. The LANDMARC is more flexible and dynamic and can achieve much more accurate and close to real-time location sensing.
As is shown in figure 1, the placement of readers and reference tags is very important
to the overall accuracy of the system obviously.
Fig1. placement of 9 readers with two different ranges and the sub-regions
The LANDMARC requires signal strength from each tag to reader. However, when this paper published, the RFID system did not provide the signal strength of tags directly to readers besides a power level distribution. As the physical distance cannot be computed accurately by using power level directly, an algorithm to reflect the relations of signal strengths by power levels is developed in this paper.
A. system setup
The prototype environment consists of a sensing network and a wireless network. The former including the RF readers and tags helps the location tracking of objects. And the latter allows wireless communication between mobile devices like PDAs and the Internet. In other words, all the tag information gathered from readers is sent over to the supplied API sitting on a specific server, and then an object’s location could be tracked after being processed in the server. Therefore, the placement of readers is unlimited.
Fig2. Siagram of the structure
B. Methodology
Alphabet table:
Here Euclidian distance Ej between a reference tag and a tracking tag is introduced. And the nearer reference tag to the tracking tag is supposed to have a smaller value.
When there are m reference tags, a tracking tag has its E vector as
Comparing all the reference results, the minimum is the nearest location. However, in the process of “comparing”, three key issues are refered through the process of locating the unknown tracking tags. I think of it as restrictions of the approach. 1) The layout of reference tags significantly affects the location accuracy. 2) How to determine the number of reference tags in a specific cell. 3) How to determine the weights assigned to different neighbors. All of them are originated from the algorithm to select the smallest value as the unknown tag’s coordinates. K-nearest-neighbor algorithm is applied here. The unknown tracking tag’ coordinates (x,y) and the weight are:
Note that the reported value of the power level is applied to take the place of the value of signal strength in the above equation.
4 Experimental results and performance evaluation
As is shown in the following diagram, a series of experiments are conducted to evaluate performance.
To quantify how well the system performs, the error distance is used as the basis for the accuracy, where n=4, m=16, u=8.
The tracking tag’s real coordinates are (x0,y0), and the computed coordinates are (x,y).
A. effect of the number of nearest neighbors
This section is used to discuss the best k value.
The coordinates of the tracking tags are computed with different k as k=1,2,3,4,5 . As is shown in Fig3, k=4 works the best. Besides, based on the statistics, it can be seen that the 50 percentile has an error distance of around 1meter while the maximum error distances are less than 2meters.
Fig3. Cumulative percentile of error distance for k from 2 to 5
B. influence of the environmental factors
This section is in order to see how well the LANDMARC system works in different environment.
- Time comparison
10 groups of data from midnight to early morning and another 10 groups of data
from 10AM to 3:00 PM are collected. Fig4 shows the comparison. There is not
much difference in the overall accuracy, which means that the LANDMARC can successfully offset the dynamics of interference.
Fig 4. Cumulative percentile of error distance in the daytime and at night - Layout comparison
The placement of the tracking tags in Fig 2a and Fig 2b is used to compare the distribution of e in both cases. Fig 5 shows that the LANDMARC helps offset environment factors that contribute to the variations in a detected range.
Fig 5. Cumulative percentile of error distance between two tracking tag placement configurations in Fig2a, Fig2b
C. effect of the number of readers
Due to the environment and the devices, one of the problems of using RF to locate objects is the inconsistency of the signal strength reception. The environment is discussed here. Among them, NLOS does create the multi-path problem and result in interference among the received signals. To better deal with it, 3 and 4 RF readers are compared respectively. As is shown in Fig 6, adding more readers may not necessary significantly increases the accuracy.
Fig 6. Cumulative percentile of error distance for 3 and 4 RF readers
D. effect of placement of reference tags
This part focuses on effect of the reference tags’ placement on accuracy.
- Higher density with the reference tags
In the next experiment, all of the reference tags with a higher density as shown
in Fig 9 (near 1) keep 48 hours. In the next 48 hours, the position of the tracking tags is changed as shown in Fig 9 (near 2). It can be seen in Fig 10 that the accuracy of the system is improved with a higher reference density in near 2.
Fig 10. Cumulative percentile of error distance with a higher reference tag density - Lower reference tags density
It is obvious that the accuracy of the LANDMARC decreases greatly with a lower
density of reference tags. Thus, there is a tradeoff between the accuracy and the number of reference tags.
Fig 11 shows two configuration of a lower reference tag density. The corresponding
distribution of error distance is shown in Fig 12.
Fig 12. Cumulative percentile of error distance with a lower reference tag density
5 Conclusion and future research
This paper has presented a prototype indoor location sensing system-LANDMARC with active RFID. Based on the statistics, it can be seen that the 50 percentile has an error distance of around 1 meter while the maximum error distances are less than 2 meters. For these experimental results, three inherent problems on RFID are discussed here. 1) None of the currently available RFID products provides the signal strength of tags directly. Notice that the subsequent RF reader has provided phase, RSSI and Doppler frequency shift for users. 2) The time interval of emitting two consecutive IDs from an active tag could not be reconfigurable. As the mechanism of avoiding signal collision, the reading interval between tags still influences the related research now. 3) From pre-tests, the power level detected by the same reader from two tags in an identical location may be different due to the variation of the behavior of tags.
Anyway, the LANDMARC proposed here brings a lot of inspiration on RFID-based indoor positioning.