Implementation of Radio Tomographic Imaging based Localisation

using a 6LoWPAN Wireless Sensor Network

Vanessa Smallbon

, Timothy Potie

, Matthew D’Souza

, Adam Postula

and Montserrat Ros

School of Information Technology and Electrical Engineering, The University of Queensland, Brisbane, Australia

School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, Wollongong, Australia

Keywords:

Wireless Sensor Networks, Radio Tomographic Imaging, Indoor Localisation.

Abstract:

Mobile localisation has numerous uses for logistics, health, sport and social networking applications. Current

wireless localisation systems typically require the use of tracking devices to be worn or implanted. The use of

tracking devices can hinder the types applications that can be used. Wireless localisation use wireless channel

propagation characteristics, such as RF receive signal strength to localise a user’s position, which requires the

use of complex radio hardware. We developed a wireless tracking system using radio tomographic imaging

to track people without wearing a mobile tracking device. We evaluated our wireless localisation network

with users in an indoor environment. Our localisation network used the 6LoWPAN wireless communications

protocol.

1 INTRODUCTION

Tracking technology has been used for many applica-

tions, including animal migration tracking and the ad-

vancement of security systems. However, traditional

tracking methods involve visual image processing or

inaccurate heat sensor technology. Both of these ap-

proaches have strong disadvantages such as the de-

pendency on light conditions for image processing

and the ﬁckle nature of sensitivity calibration for in-

frared sensors. Radio Tomographic Imaging (RTI)

uses inexpensive radios to track objects in a closed

area without the need for radio tags. It exploits the

ability of radio waves to travel through objects such

as trees and walls.

Current localisation techniques depend on using

sensing infrastructure already present in the environ-

ment such as visual markers, wireless LAN hotspots,

cellular networks or GPS satellite coverage. RF lo-

calisation methods such as Received Signal Strength

Indicator (RSSI) or Time of Arrival also experience

inaccuracies and reliability issues when operating in-

doors.

Current wireless localisation systems typically re-

quire the use of tracking devices to be worn or im-

planted. The use of tracking devices can hinder the

types applications that can be used. Current position

localisation use wireless channel propagation charac-

teristics, such as RF receive signal strength to localise

a user’s position, which requires the use of complex

radio hardware.

We developed a wireless tracking system that used

RSSI radio tomography to track people and objects

without using a mobile tracking device. Our wire-

less localisation system used a low powered wire-

less sensor network infrastructure which consisted of

reference nodes placed at predetermined coordinates.

The network of reference nodes continuously mea-

sure RSSI and link quality parameters. These parame-

ters are then used to form a tomographic image, show-

ing the locations of signiﬁcant RF signal attenuation.

RF signals that pass through the human body or

solid inanimate objects are subjected to attenuation

RF signals and hence will affect the RSSI and link

quality between pairs of reference nodes. Locations

of RF signal attenuation are then used to determine

the position of a user or object. We developed a Ra-

dio Tomographic Localisation System (RTLS) that lo-

calises the positions of people and objects. We evalu-

ated the RTLM in a typical and realistic indoor envi-

ronment.

This article is organised into the following sec-

tions. Section 3 presents an overview of the wireless

localisation network infrastructure used. Section 4

describes the operation of the RTLM. An evaluation

of the RTLM is discussed in section 5. Conclusions

and further work are presented in section 6.

Smallbon V., Potie T., D’Souza M., Postula A. and Ros M..

Implementation of Radio Tomographic Imaging based Localisation using a 6LoWPAN Wireless Sensor Network.

DOI: 10.5220/0005513400270032

In Proceedings of the 12th International Conference on Wireless Information Networks and Systems (WINSYS-2015), pages 27-32

ISBN: 978-989-758-119-9

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

2 LITERATURE REVIEW

Different types of wireless technologies have been

investigated for indoor location systems. One such

approach is the use of dead-reckoning, such as that

by Klingbeil et al (Klingbeil and Wark, 2008) who

developed an indoor localisation system using dead-

reckoning with a hip-worn mobile node that detected

a user’s footsteps and heading and used a particle ﬁl-

tering process to estimate the position of the user.

Widyawan et al (Widyawan et al., 2008) also designed

a dead-reckoning system that used a foot-mounted in-

ertial sensor for heading and footstep detection com-

bined with a backtrack particle ﬁltering process. Al-

though dead-reckoning can achieve a sufﬁcient accu-

racy, the disadvantage of such systems is the require-

ment of users to wear a mobile device.

Wilson et al (Wilson and Patwari, 2010) devel-

oped an RTI system for indoor tracking, using a wire-

less sensor network. Zhao et al (Zhao et al., 2013)

used a kernel distance based estimation method for

estimating link quality for RTI. This was found to be

suitable for detecting moving people. We used a sim-

ilar approach. Qiu et al (Qiu et al., 2010) and Hu et

al (Hu et al., 2014) investigated the use of machine

learning techniques for RTI, in order to reduce inac-

curacies caused by noisy RSSI measurements.

Bonior et al (Bonior et al., 2015) developed an

RTI system implemented using software radios, in or-

der to validate the accuracy of using RSSI measure-

ments for RTI. Wang et al (Wang et al., 2015) used,

a Variational Bayesian Gaussian mixture model and

K-means clustering to improve object tracking in an

RTI system. Amendolare et al (Amendolare et al.,

2014) developed an RTI system that used both static

and mobile reference nodes for indoor environments

in ﬁrst responder scenarios Martin et al (Martin, 2015;

Martin et al., 2014) presented a beam forming based

RTI model in order to improve position accuracy and

to reduce the image frame rate latency.

3 WIRELESS REFERENCE NODE

NETWORK

The wireless reference node network was based on

typical wireless sensor network infrastructure. Wire-

less sensor networks are used for a sensing and ac-

tuation applications. Wireless sensor network infras-

tructure are used for low powered indoor and outdoor

based applications and are designed to be portable and

easy to deploy, compared to other wireless LAN net-

work infrastructure. used to provide realtime received

signal strength measurements. The network consisted

of wireless reference nodes placed around the track-

ing zone. The base node was placed outside the track-

ing zone. The tinyOS 6LoWPAN based BLIP com-

munications protocol was used by the wireless refer-

ence node network.

Figure 1: Overview of Wireless Reference Node Network.

The wireless reference node network as seen in

Figure 1 consisted of two types of nodes: base and

reference nodes. The reference nodes are placed

around the boundaries of the zone in which users

are tracked in. The reference nodes are used to

measure the radio received signal strength from the

base node. The server connected to the node node

displays the current position of the person. The

reference and base nodes used the Zigduino plat-

form (Logos Electromechanical LLC, 2013) with

TinyOS (TinyOS, 2013). The Zigduino uses the At-

mega128RFA1 Wireless System on Chip (SoC) that

has an Atmega128 microcontroller and a 2.4GHz Zig-

bee/802.15.4 transceiver (Atmel Corporation, 2012).

The 6LoWPAN protocol was used to provide a wire-

less communication link between the base and refer-

ence nodes.

3.1 Reference Node

Figure 2: Reference Node.

The reference node, seen in Figure 2, communi-

cates to the base node using a 6LoWPAN network

connection. The reference nodes are by the RTLM

to locate people and objects in the tracking zone. The

position of each reference node is known by the base

node. Each reference node has a predetermined ID

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems

which is included in each packet transmitted. The

reference node is powered by a rechargeable battery.

As mentioned previously, the reference node is im-

plemented using the Zigduino platform with the At-

mega128RFA1 SoC. The Atmega128RFA1 has a re-

ceiver sensitivity range of 88dB with 1dB resolution

and has a minimum signal power detection threshold

of -90dBm (Atmel Corporation, 2012). The transmis-

sion power can vary from 3.5dBm to -16.5dBm. The

reference node used a transmission power setting of

-16dBm.

3.2 Base Node

Figure 3: Base Node.

The base node, seen in Figure 3, receives and pro-

cesses packets from each reference node. The base

node is connected by a USB connection to a server

computer. The server computer tracks the position

of the person using the RTLM. The server com-

puter used is a Raspberry Pi embedded linux plat-

form (Raspberry Pi Foundation, 2013). A graphical

user interface, written in Python was used to display

the current status of the reference nodes (transmitting

or receiving) and the heatmap. This is shown in Fig-

ure 4.

Figure 4: Graphical User Interface showing Heatmap and

Node Status.

3.3 Reference Node Protocol

The Reference Node Protocol (RNP) was developed

to provide maximum received signal strength resolu-

tion required by the RTLM. The RNP was used by

each reference node to measure the RSSI and to syn-

chronise the transmission of each reference node. The

protocol was required to minimise radio interference

and to be self regulating. RNP was implemented us-

ing 6LoWPAN with the TinyOS BLIP library. RNP

used the IP protocol User Datagram Protocol (UDP)

to transfer packets between the nodes.

The RNP packet can be seen in Figure 5, consists

of a sender ID and the RSSI values of neighbouring

reference nodes detected. RNP uses a round robin

slot scheme, in which each reference node will only

transmit, depending the reference sender node ID last

received. The RNP used a slot time period of 300ms.

Figure 5: Reference Node Protocol Packet Format.

4 RADIO TOMOGRAPHIC

LOCALISATION SYSTEM

The Radio Tomographic Localisation System (RTLS)

estimates a person’s position using RSSI attenuation

measurements. RSSI is the measured received sig-

nal strength between a transmitter/receiver reference

node link. RF signals that pass through the human

body or inanimate objects are subjected to attenua-

tion or reﬂection, which affects the RSSI and between

a transmitting and receiving reference node link. The

RTLM uses the network of reference nodes to con-

tinuously measure RSSI and link quality parameters.

The RTLM forms a tomographic image of RF sig-

nal strength and determines the locations of signiﬁ-

cant RF attenuation. Locations of attenuation are then

used to determine the position of a person in the track-

ing zone.

Figure 6 shows an overview of the RTLM. The

RTLM was based on kernel distance estimation as

used by Zhao et al (Zhao et al., 2013). The RTLM

ﬁrst estimates the attenuation between pairs of refer-

ence nodes using RSSI histogram ﬁltering and link

attenuation estimation. Once the link quality between

the reference node links has been computed, a tomo-

graphic image showing the heatmap of signal attenu-

ation can be formed. Areas of signiﬁcant attenuation

are then used to determine the location of objects and

users. A grid averaging is then used to reduce dis-

tortions in the tomographic image, caused by ﬂuctua-

tions in link attenuation estimates.

ImplementationofRadioTomographicImagingbasedLocalisationusinga6LoWPANWirelessSensorNetwork

Figure 6: Radio Tomographic Localisation System

Overview.

4.1 RSSI Histogram Filtering

The RSSI readings for a link between reference nodes

is used to determine the level of attenuation between

the reference nodes. For each link, two histograms

were created to track the distribution of RSSI values.

The histogram is used to measure the occurrence of

RSSI values. For each link, a short and long time

histogram was used. The short term histogram keeps

track of the most recent RSSI values while the long

term histogram tracks the long RSSI values over a

longer period of time. The use of two histograms

is advantageous as it allows dynamically moving ob-

jects to be detected with more certainty as discussed

in (Zhao et al., 2013).

4.2 Link Attenuation Estimation

The link attenuation is used to determine if a person

or object is between the nodes. For each link, the link

attenuation is calculated using the difference between

the long and short term histograms. The difference

between the long and short term histograms is a mea-

sure of the change of attenuation that has occurred for

a particular link. The kernel distance is used to cal-

culate the difference between the long and short term

histograms. The change in attenuation is then used to

form a tomographic image.

4.3 Grid Averaging

Link attenuation can ﬂuctuate signiﬁcantly, which can

cause distortions in the tomographic image. In order

to reduce ﬂuctuations in link attenuation, a grid aver-

aging method was used. The tracking zone is divided

into an evenly spaced grid, as shown in Figure 7. The

link attenuation estimates of each link that crosses

a particular grid are averaged and then assigned the

same averaged value.

Figure 7: Grid Averaging.

4.4 Weighted Elliptical Model and

Tomographic Image

A tomographic image of the tracking zone is formed

using the link attenuation estimates. A weighted el-

liptical model using the link attenuation estimates is

formed, as shown in Figure 8. Using the a wighted

elliptical model, an image can be formed where each

pixel is assigned a value, if it falls within the ellipse

formed by the reference nodes in the link. Each pixel

value corresponds to a heat map colour value.

Figure 8: Example of Weighted Elliptical Model using Ref-

erence Node IDs.

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems

5 EVALUATION

The RTLM was tested in an indoor environment

shown in Figure 9. The reference node network was

setup in a 5m by 5m open space. We tested the RTLM

by having a person stand and walk, within the track-

ing zone. Figure 10 shows the person standing in the

middle of the lower boundary. Figure 11 shows the

person standing in the middle of the right boundary.

Figure 12 shows the person in the top right corner of

the tracking zone.

Figure 9: Reference node Localisation Network Deploy-

ment.

Figure 10: Person Standing in the Middle of the Lower

Boundary.

6 CONCLUSIONS

We presented a wireless indoor localisation system

that tracked people in an indoor environment, using

radio tomographic imaging. We developed the RTLM

which formed a tomographic image using RSSI link

quality estimation. The RTLM used a 6LoWPAN

based wireless sensor network, to measure link atten-

uation between a pair of reference nodes. The link

attenuation was then used to form a tomographic im-

Figure 11: Person Standing in the Middle of the Right

Boundary.

Figure 12: Obstacles in the upper left boundary.

age, showing the locations of signiﬁcant RF signal at-

tenuation.

Locations of RF signal attenuation are then used

to determine the position of a user or object. We eval-

uated the RTLM in an realistic indoor environment.

The RTLM was able to detect people and obstacles in

the tracking zone. Further work involves, developing

a self power calibration mechanism, allow for multi-

ple people tracking and to track users in large indoor

ofﬁce environments. We are also investigating the use

of a tablet interface to view the heatmap and tracking

information.

ACKNOWLEDGEMENTS

The authors would like to thank the School of ITEE

at The University of Queensland and the Pervasive

Computing Group at CSIRO for their support in this

project.

REFERENCES

Amendolare, V., Cyganski, D., and Duckworth, R. (2014).

Transactional array reconciliation tomography for

ImplementationofRadioTomographicImagingbasedLocalisationusinga6LoWPANWirelessSensorNetwork

precision indoor location. Aerospace and Electronic

Systems, IEEE Transactions on, 50(1):17–32.

Atmel Corporation (2012). Atmega128rfa1 datasheet.

Bonior, J., Hu, Z., Guo, T., Qiu, R., Browning, J.,

and Wicks, M. (2015). Software-deﬁned-radio-based

wireless tomography: Experimental demonstration

and veriﬁcation. Geoscience and Remote Sensing Let-

ters, IEEE, 12(1):175–179.

Hu, Z., Hou, S., Wicks, M., and Qiu, R. (2014). Wireless to-

mography in noisy environments using machine learn-

ing. Geoscience and Remote Sensing, IEEE Transac-

tions on, 52(2):956–966.

Klingbeil, L. and Wark, T. (2008). A wireless sensor

network for real-time indoor localisation and motion

monitoring. In Information Processing in Sensor Net-

works, 2008. IPSN ’08. International Conference on,

pages 39 –50.

Logos Electromechanical LLC (2013). Zigduino R2.

”http://logos-electro.com/zigduino/”.

Martin, R. (2015). Inverse beamforming for radio tomogra-

phy. Signal Processing Letters, IEEE, 22(2):187–191.

Martin, R., Folkerts, A., and Heinl, T. (2014). Accuracy

vs. resolution in radio tomography. Signal Processing,

IEEE Transactions on, 62(10):2480–2491.

Qiu, R., Hu, Z., Wicks, M., Hou, S., Li, L., and Gary, J. L.

(2010). Wireless tomography, part ii: A system engi-

neering approach. In Waveform Diversity and Design

Conference (WDD), 2010 International, pages 277–

282.

Raspberry Pi Foundation (2013). Raspberry Pi.

”http://www.raspberrypi.org/”.

TinyOS (2013). TinyOS. ”http://www.tinyos.net/”.

Wang, Q., Yigitler, H., Jantti, R., and Huang, X. (2015).

Localizing multiple objects using radio tomographic

imaging technology. Vehicular Technology, IEEE

Transactions on, PP(99):1–1.

Widyawan, Klepal, M., and Beauregard, S. (2008). A

backtracking particle ﬁlter for fusing building plans

with pdr displacement estimates. In Positioning, Nav-

igation and Communication, 2008. WPNC 2008. 5th

Workshop on, pages 207 –212.

Wilson, J. and Patwari, N. (2010). Radio tomographic imag-

ing with wireless networks. Mobile Computing, IEEE

Transactions on, 9(5):621–632.

Zhao, Y., Patwari, N., Phillips, J. M., and Venkatasubra-

manian, S. (2013). Radio tomographic imaging and

tracking of stationary and moving people via kernel

distance. In Proceedings of the 12th international

conference on Information processing in sensor net-

works, IPSN ’13, pages 229–240, New York, NY,

USA. ACM.

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems