Evaluation of Range-based Methods for Localization in Grain Storages

Jakob Pilegaard Juul

1,2

, Ole Green

and Rune Hylsberg Jacobsen

Department of Engineering, Aarhus University, Aarhus, Denmark

Kongskilde Strategic Development, Kongskilde Industries A/S, Sorø, Denmark

Keywords:

Wireless Sensor Networks, WSN, Localization, Biomass.

Abstract:

Monitoring biomass storages by using wireless sensor networks with localization capabilities can help prevent

economic losses during storage, help to improve the grain quality and lower costs during drying. In this article,

the received signal strength was used to perform localization of wireless sensor nodes embedded in a grain

storage. A path loss model that takes into account the temperature and moisture content of the grain at each

sensor node was used for estimating distance based on received signal strength. The average error of the

position estimates was 6.3 m. Tests using near-ﬁeld electromagnetic ranging were performed to evaluate the

performance of the method. It was found that the experimental setup worked best between 2 - 7 m where the

average error was 4.9% of the actual distance.

1 INTRODUCTION

Many wireless sensor network (WSN) applications

can beneﬁt from nodes knowing their location. This

could be for use in identiﬁcation and correlation

of gathered data, node addressing, geographic rout-

ing, object tracking, etc. (Boukerche et al., 2007).

Biomass is biologically active or potentially active

given the right conditions. When kept in a storage,

this biological activity is unwanted and may result in

a degradation of the quality of biomass. The com-

bination of continuous monitoring and localization

provides the information needed to limit the biomass

degradation. In the case of grains and seeds (grain

from here on), storage is usually preceded by a drying

process to lower the moisture content of the grain to

a level that ensures minimal biological activity while

still providing a good quality crop for the intended

use. The main expenses of the drying process are the

cost of fuel and electrical power for heaters and fans

(Mühlbauer, 1986). By using a WSN with localiza-

tion capability, it enables the user to see the mois-

ture content from different sections of the grain and

only dry the sections that have not reached the de-

sired moisture content. This lowers the total amount

of fuel and electricity used as the system only dries the

sections in need and not the entire storage throughout

the whole process. This sectioning of the storage dur-

ing the drying process also helps to avoid overdrying.

Overdrying happens when some sections of a storage

have reached the desired moisture content while other

sections have not. The opposite may also happen,

where the system stops before the moisture content

of all the grain in the storage has been lowered to a

stable level. This can result in losses during the stor-

age period due to high biological activity.

This paper presents the experimental results of us-

ing radio signal strength (RSS) localization for posi-

tioning of wireless sensor nodes embedded in a grain

storage. It also examines the potential of using a

near-ﬁeld electromagnetic ranging (NFER) technique

in addition to RSS for localization of wireless nodes.

Section 2 describes related work. Section 3 describes

the theory behind the RSS localization used in this

study and the theory behind NFER. Section 4 de-

scribes the experiments and results. Section 5 dis-

cusses the results of the experiments and Section 6

concludes and outlines future work.

2 LOCALIZATION METHODS

Localization in WSNs can be carried out by using

various methods. However, many localization algo-

rithms implement the same basic techniques. Those

algorithms are based on ﬁnding the distance between

nodes, deriving a position based on those distance

measurements and lastly reﬁne the position estimate

based on the distance to other nodes (Langendoen and

Reijers, 2003). To localize a node in a WSN using this

technique the literature divides different approaches

Juul, J., Green, O. and Jacobsen, R.

Evaluation of Range-based Methods for Localization in Grain Storages.

DOI: 10.5220/0005669601050112

In Proceedings of the 5th International Confererence on Sensor Networks (SENSORNETS 2016), pages 105-112

ISBN: 978-989-758-169-4

105

into range-based and range-free methods, see Fig. 1.

Range-based techniques use distance estimates or an-

gle estimates in location calculations while range-free

techniques depend only on the contents of received

messages (Hu and Evans, 2004).

Localization methods

Range-based

Range-free

RSS TOA AOATDOA

Proximity

Hop

count

NFER

Figure 1: Overview of localization methods.

Proximity methods are among the range-free lo-

calization techniques that provide a coarse localiza-

tion of a node. These methods do not measure dis-

tance directly, rather they rely on the number of node-

hops required to reach a destination node (Kumar and

Lobiyal, 2013) or on the connectivity by determining

which ﬁxed-position anchors are within range. As this

approach requires no special equipment, the use cases

are many, especially in WSNs with many nodes. It has

been demonstrated that by assuming a sensor node

was located on the centroid between anchors in range,

it could be located with an error of less than 30%

of the distance between the beacons (Bulusu et al.,

2000). However, the presence of multipathing, noisy

environments, dead spots, and fading, etc. is likely to

reduce the accuracy of the localization.

RSS has been used both as a range-free and a

range-based method. An advantage of RSS-based

methods is that almost any radio chip supports RSS

measurements. It is possible to translate RSS to range

using a path loss model. The model should be cho-

sen with care as it is very sensitive to the environ-

ment where the node is deployed. It has been in-

vestigated how well RSS performs in real situations

(Whitehouse et al., 2007). Several parameters such

as attenuation coefﬁcient and effective range of com-

munication were characterized by using a large num-

ber of wireless nodes in different settings. Comparing

the locations estimated from RSS measurements of 49

nodes to GPS data, standard error in location of 4 m

was achieved.

Time-of-arrival (TOA) has also been used for lo-

calization (Gezici and Poor, 2009). A well-known ex-

ample of a one-way TOA system is the Global Posi-

tioning System (GPS). The one-way TOA approach

requires stringent clock synchronization between the

nodes and low-jitter clocks on the nodes (Gezici and

Poor, 2009). Two-way TOA is also possible. In this

case, no clock synchronization is required between

nodes. Instead, a delay-calibration can be performed

where the delay between reception and response to a

message is recorded. A study has shown that TOA

localization techniques can estimate the location of

a node with approximately 15 cm accuracy (Sathyan

et al., 2011). Higher accuracy is possible by using a

more bandwidth.

Time-difference-of-arrival (TDOA) is a variation of

TOA where measurements of the time difference be-

tween the arrival of two or more signals facilitate lo-

calization. The signals can be from a single transmit-

ter using signals with different propagation speeds,

e.g. sound and radio signals, or it can be of the same

type of signals transmitted from/received at multiple

locations. An example of the ﬁrst case is small units

with ultrasonic transducers and radios as in Cricket

localization-support system (Priyantha et al., 2000).

The main drawbacks of such a system are the limited

propagation range of the ultrasound wave and the po-

tential interference problems with other ultrasounds

sources in outdoor environments (Bulusu et al., 2000).

TDOA does not require the same stringent synchro-

nization as TOA between transmitter and receiver but

clock skew and drift has to be taken into account

(Xiong et al., 2015).

Angle-of-arrival (AOA) can also be used to determine

location. Traditionally this is done using phase inter-

ferometry techniques and large antenna arrays. AOA

has demonstrated good performance in high signal-to-

noise ratio (SNR) conditions but require large anten-

nas and fail in the presence of multipathing and co-

channel interference (Rappaport et al., 1996). A dif-

ferent approach more suited for WSN is to use smart

antennas for AOA estimation. These are smaller than

conventional antenna arrays and can be used with

high-resolution AOA estimation algorithms such as

MUSIC or ESPRIT (AlHajri et al., 2015). Apply-

ing the AOA approach to WSNs means that the nodes

and anchors with AOA capabilities require special

equipment, usually in the form of special antennas

(Nasipuri and Li, 2002).

NFER relies on measuring the difference in phase be-

tween the electric and magnetic components in the

near ﬁeld of a signal. This difference can be related

to the distance to the transmitter. Shantz et al. has

successfully used NFER for localization and reported

average position errors of 30 cm. It was noted that

the error had little correlation with range but that de-

creasing SNR could severely affect the phase accu-

racy (Schantz, 2005).

SENSORNETS 2016 - 5th International Conference on Sensor Networks

106

(a) (b)

Figure 2: The permittivity of ’Nebred’ hard red wheat at 24

◦

C and the indicated moisture content. The density of the wheat

was 768 kg/m

at 13% moisture content. (a) shows the relative permittivity over a range of moisture content, and (b) shows

the loss factor. Graph recreated from (Nelson, 1965).

3 RSS AND NFER

LOCALIZATION THEORY

The theory for RSS localization as well as the theory

behind NFER is presented in this section.

3.1 Received Signal Strength

To use RSS in localization a model that relates RSS

to distance has to be found. The model used in this

paper to estimate the distance between a transmitter

and a receiver is the log-normal path loss model as

shown in Eq. (1) (Rappaport, 2002).

PL(d)[dBm] = PL(d

)[dBm] − 10 · n · log(

) + X

(1)

Here PL(d) is the average received power at distance

d, PL(d

) is the average measured received power at

the reference distance d

. X

is a zero-mean Gaussian

random variable with standard deviation σ which ac-

counts for the effect of shadowing (Rappaport, 2002).

n is the path loss exponent. In free space the exponent

is equal to two. In the case where the sensor nodes

are embedded in biomass the exponent is higher. As

can be seen in Fig. 2 the complex relative permittiv-

ity of the grain is affected by the temperature, mois-

ture content, and frequency of the signal. This means

that all these three parameters will affect the path loss

of the transmitted signal. The frequency-dependent

complex relative permittivity is deﬁned as Eq. (2).

(ω) = ε

(ω) + jε

(ω) (2)

Where ω is the angular frequency, ε

(ω) relates to the

electrical energy that can be stored in the medium.

(ω) is the imaginary part of the complex relative

permittivity and relates to the dissipation of energy in

a material. To account for the moisture content and

temperature of the grain, n is deﬁned as:

n = a + b · M

+ c · T (3)

In Eq. (3), M

is the moisture content wet-basis of

the grain. The moisture content is found by using

the modiﬁed Henderson equation (ASAE Standards,

1999). T is the temperature in degrees Celcius, a, b

and c are the ﬁtting parameters. a, b and c can be

found by using linear regression given measured RSS

values at known distances.

Once the distance has been estimated these are to

be used for 3-dimensional position estimates by using

trilateration. In this case, it is the 3-dimensional po-

sition that is of interest and the trilateration problem

resolves into ﬁnding the intersections of three spheres.

As this gives none, one, or two solutions, knowledge

of the placement of the anchors and sensor nodes can

be used to eliminate solutions in the cases where more

than one is found.

3.2 Near-ﬁeld Electromagnetic Ranging

The near ﬁeld extends to about half a wavelength λ

from an electrically small antenna (Schantz, 2005)

and is governed by other phenomena than those of

the far-ﬁeld region. In the far-ﬁeld region the phase

difference between the electric and magnetic compo-

nents is constant. In contrast, in the near-ﬁeld region

of an antenna with dimensions smaller than 1/10 of

the wavelength, also called an electrically small an-

tenna (Miron, 2006), the phase of the two components

change with distance. If each of these components are

measured separately and compared to each other, this

can be used as a measurement of distance. The re-

lation between distance d and phase difference ∆φ is

Evaluation of Range-based Methods for Localization in Grain Storages

107

given by Eq. (4).

d =

2π

cot∆φ (4)

Here d is the distance from between transmitter and

receiver. λ is the wavelength of the signal. ∆φ is

the difference in phase between the electric and mag-

netic component of the radio wave. In practice, the

effect can be used out to around 0.25 λ depending

on the transmitted power (Kim et al., 2006). Fig. 3

shows how the phase difference changes as a function

of the distance to between transmitter and receiver in

fractions of the signal wavelength. NFER has sev-

Figure 3: Plot showing the difference in phase as a result

of distance from the transmitter. The interval between the

vertical lines indicates the best area of operation.

eral interesting characteristics. The measurement is

one-way, meaning that only the receiver needs to do

the measurement, and it is not required to communi-

cate with the transmitter. The accuracy of the mea-

surements is not affected by the available bandwidth,

making the method suitable for situations where high

frequency signals are difﬁcult to use as in the case of

biomass. For the phenomenon to be of use at ranges

from tens to hundreds of meters frequencies between

500 kHz (λ = 600 m) to 7 MHz (λ = 42.8 m) are ap-

propriate. Because low frequencies tend to be more

resistant to multipath effects and have better penetra-

tion depth, this is another advantage of the approach.

The exact choice of frequency should be done with

care. The best area of operation is between 0.08λ and

0.25λ (Kim et al., 2006).

4 EXPERIMENTAL SETUP AND

RESULTS

To examine the use of RSS and NFER for localiza-

tion in biomass two experiments were performed. The

ﬁrst experiment examined the performance of a sys-

Figure 4: The placement of sensor nodes and base stations

during the experiment. The spheres are sensors and the

boxes are base stations.

tem based solely on RSS measurements for localiza-

tion. The second experiment examined the ranging

precision of a NFER setup.

4.1 RSS Measurements

The goal of the experiment was to ﬁnd the 3-

dimensional position estimates of a number of nodes

deployed at known positions in a grain storage. This

experiment was carried out from August 2011 to

November 2011.

In the experiment 30 nodes and 3 base stations

were deployed in known positions in a wheat stor-

age as shown in Fig. 4. The dimensions of the wheat

storage was 9.6m · 15.8m· 4.0m The nodes used in the

experiment consist of a MSP430 microcontroller with

55 kB onboard ﬂash and 5 kB of RAM, 1 MB exter-

nal ﬂash memory, a Semtech SX1231 radio, a 8.5 Ah

lithium C-cell battery and a SHT11 temperature and

relative humidity sensor. The nodes run TinyOS

(Levis et al., 2005) and communicate at 433 MHz.

For a more in-depth description of the system, please

refer to (Juul et al., 2015).

The nodes measured and reported temperature and

humidity every 5 minutes. During each transmission

the RSS of the signal was recorded. The RSS mea-

surements were averaged over 30 minute intervals and

were subsequently used to try and estimate the dis-

tance from each base station to each individual sen-

sor. This estimate was used to make a 3D-position

estimate that was compared to the recorded position

of the nodes. During the RSS-experiment data was re-

ceived from 18 out of 30 nodes. 6 of these 18 dropped

out before the end of the experiment but did deliver

data for a limited period. The cause of the lack of

reporting nodes was drainage of power and transmis-

sion loss. The missing nodes can be seen as gray dots

in Fig. 4.

The temperature and calculated moisture content

of the remaining nodes can be seen in Fig. 5 and 6.

The graphs show the different conditions present in

the storage and how it changes over time. The value

SENSORNETS 2016 - 5th International Conference on Sensor Networks

108

Figure 5: The temperature measured by the nodes.

Figure 6: The moisture content wet-basis of the grain.

Calculated using the modiﬁed Henderson equation (ASAE

Standards, 1999).

of the ﬁtting parameters in Eq. (3) can be found by

using the measured RSS values for each node to each

base station and doing least squares linear regression.

The values found were a = 7.028, b = -0.082, and c

= -0.016. A comparison of the actual and estimated

distances can be seen in Fig. 7. The average error of

the estimated distance between a base station and a

node is 33.54% of the actual distance between them.

The average error in meter for all measurements is

2.67 m.

Using trilateration as outlined earlier to estimate

positions from the range estimates gives an average

positioning error of 6.3 m.

4.2 Near-ﬁeld Phase Measurements

In an effort to ﬁnd a distance-based measurement to

supplement RSS, work was done to study the feasi-

bility of the NFER method. The purpose of the ex-

periment was to measure the difference in phase be-

tween the electric and magnetic component of the ra-

Figure 7: The estimated distance between the individual

nodes and the base stations compared to the actual distance.

7.03MHz

bandpass

filter

7.03MHz

bandpass

filter

USRP GNU Radio

Low-noise

amplifier

Low-noise

amplifier

Whip antenna

Loop antenna

Whip antenna

7.03MHz CW radio

transmitter

Figure 8: Diagram of the experimental setup.

dio wave at ranges of 1 to 10 m. To get the biggest

change in phase at the desired range a frequency of

7.03 MHz (λ = 42.6 m) was used. Three electrically

small antennas were constructed for the experiment.

Two whip antennas and a box loop antenna. The

whip antennas were matched to the feedline by an air

core inductor and the loop by an adjustable capacitor

with a range of 15 - 385 pF. A small battery powered

7.03 MHz Continuous Wave (CW) radio connected to

one of the whip antennas was used as the transmitter.

The receiver consisted of a N210 USRP from Ettus

that had a whip and loop antenna connected to the two

inputs of the device. Between antennas and the USRP

was placed a ZFL-1000LN+ low-noise ampliﬁer fol-

lowed by a Bessel bandpass ﬁlter with a center fre-

quency of 7 MHz and a passband of 500 kHz. Fig. 8

shows the experimental setup. The experiment took

place on a ﬁeld with line of sight between transmitter

and receiver. Measurements were taken at 1 to 10 m

in 1 m intervals. The transmitter was the mobile unit

while the receiver was in a ﬁxed position for all mea-

surements. By using GNU Radio version 3.7.6, the

received signal was ﬁltered and stored in a ﬁle for of-

ﬂine processing. Each antenna added a constant phase

shift to the measurements as they were not tuned ex-

actly to resonance due to the limitations of the tuning

circuit. The phase shift of an antenna was subtracted

from the measurements made by that antenna and the

Evaluation of Range-based Methods for Localization in Grain Storages

109

Figure 9: Theoretical and measured phase difference be-

tween magnetic and electric component of a 7.03 MHz sig-

nal in the near-ﬁeld.

Figure 10: The range estimated from the phase difference.

Figure 11: The estimation error at a given distance.

phase difference between the two signals was calcu-

lated. Fig. 9 shows the calculated phase difference,

Fig. 10 shows the range estimates and Fig. 11 shows

the estimation error at a given range.

In Fig. 9 the measured phase differences are over-

laid on the theoretical phase difference at a given

range. Up to approximately 7 m the curve of the the-

ory and the curves of the measurements series ﬁt to-

gether well. Above 7 m the measurements and theory

increasingly diverge. This can also be seen in the dis-

tance estimates in Fig. 10. From 2 m to 7 m the mea-

surements have an average error of 21 cm, and the av-

erage error at a given distance is 4.9% of the distance,

see in Fig. 11. At 1 m the average error is 71 cm.

This is much larger than the immediately following

measurements. At this range the antenna dimensions

become a signiﬁcant proportion of the range and the

theoretical prediction breaks down (Schantz, 2005).

Looking at Fig. 11 also show that the error increases

with range. This is because of the decreasing SNR

as the distance between transmitter and receiver in-

crease. The average SNR at 2 m was 57 dB and the

SNR at 7 m was 28 dB.

It should be noted that the NFER measurements

were done in air. When this method is applied in

biomass, the increased RF absorption will lower the

SNR at all ranges. The heterogeneity of the material

would also contribute to measurement errors.

5 DISCUSSION

Considering the dimensions of the grain storage, a

positioning error of 6.3 m for the RSS trilateration

is large. Looking at the actual position estimates

roughly 50% of the estimates are placed outside the

storage due to estimation errors. In this experiment,

only three base stations were used. Adding more

bases in known positions would likely improve the es-

timates (Alippi and Vanini, 2006). Improving the path

loss model used could also reduce error. The current

model includes the temperature and moisture content

at the node, but these parameters change as the signal

travels through the grain, see Fig. 6. If this hetero-

geneity is included in the path loss model improved

location estimates are expected. The same holds true

if the fact that signals from different nodes travel dif-

ferent distances through the grain is included in the

model.

At ranges from 2 to 7 m the current NFER-setup

provides a range estimate with an average error of

4.9% of the actual distance and shows to be primarily

limited by the SNR. This indicates that if sufﬁcient

signal is present measuring the near ﬁeld phase has

the potential of providing accurate and precise dis-

tance estimates. On the other hand making electri-

cally small and efﬁcient antennas is not trivial and is

the biggest difﬁculty in trying to increase range and

lower the error.

Table 1 compares the NFER experiment carried

out here with experiments carried out with other rang-

SENSORNETS 2016 - 5th International Conference on Sensor Networks

110

Table 1: Comparing different ranging methods.

Name of method Range Accuracy Suitability for biomass

NFER 7 m 20 cm Partly

NFER (Schantz, 2005) 70 m 30 cm Partly

RSS (Whitehouse et al., 2007) 50 m 4 m Partly

TOA (Sathyan et al., 2011) 30 m 15 cm No

TDOA, ultrasound (Balakrishnan et al., 2003) 10 m 1 cm No

AOA (Amundson et al., 2010) - 3

◦

ing methods. The range NFER operates in is deter-

mined by the frequency of the signal used. In the cur-

rent experiment the range was limited to below 10 m

while Schantz et. al. (Schantz, 2005) used a fre-

quency of 1.295 MHz to achieve a longer range. The

low frequencies used in NFER gives a higher pene-

tration depth in materials. However, while NFER al-

ways requires an electrically small antenna, the lim-

ited space in the sensor nodes puts even shaper con-

straints on the size. Because of this it is only partly

suitable for use in biomass.

As RSS-measurements can be made on any re-

ceived signal, the approach is possible to use in

biomass as long as the network can communicate. It

should be noted that modeling the path loss is difﬁ-

cult even in good condition and the biomass makes

this even more difﬁcult to make accurate estimations.

However, due to the high availability of the measure-

ment even in this environment it is deemed partly suit-

able for use in biomass.

TOA require high bandwidth to provide a good

accuracy (Thorbjornsen et al., 2010). The high-

frequency signals needed for high bandwidth are

severely attenuated by obstacles. As the nodes are to

be embedded in biomass, TOA is not a suitable local-

ization approach.

As shown in Table 1, TDOA can have a very low

error. However, in the case of ultrasound TDOA, the

signal has a short range in air and it would be further

attenuated in biomass. Having a radio signal recorded

by multiple receivers suffer from the same bandwidth

limitations as TOA in biomass.

AOA is not suitable for use in biomass. The an-

tennas needed are too large and, while beamform-

ing has no size requirements, the heterogeneity of

the biomass would change the beam pattern unpre-

dictably, making it difﬁcult to get accurate AOA mea-

surements.

Based on the measurements presented here, RSS

localization seems poorly suited as a stand-alone sys-

tem for estimating the position of wireless sensor

nodes in grain. The experiments with the NFER-

approach looks promising, but as measurements have

only been done in air further experiments to test the

performance in grain should be carried out. Com-

bining the two methods for localization in biomass

should be investigated. The use of data fusion tech-

niques such as weighted least squares and Kalman ﬁl-

ters could improve position estimates compared to us-

ing only one of the methods.

6 CONCLUSIONS

This paper examines the use of RSS localization in a

grain storage and NFER as a ranging method. The

RSS localization had an average positioning error of

6.3 m. To minimize the error, it is suggested to use

more base stations at known positions and put more

effort into reﬁning the path loss model used. Investi-

gations were made on using NFER for localization. A

ﬁrst experiment tested NFER localization in air, and

the feasible distance of the setup was limited to dis-

tances between 2 and 7 m. These estimates had an

average error of 4.9% of the actual distance.

Future work includes improving the antenna and

RF circuit design to improve the SNR and then test

NFER in a grain storage. Furthermore, a system com-

bining RSS and NFER for localization should be im-

plemented and tested in a grain storage.

REFERENCES

AlHajri, M. I., Goian, A., Darweesh, M., AlMemari, R.,

Shubair, R. M., Weruaga, L., and Kulaib, A. R. (2015).

Hybrid RSS-DOA Technique for Enhanced WSN Lo-

calization in a Correlated Environment. In 2015 Inter-

national Conference on Information and Communica-

tion Technology Research (ICTRC), pages 238–241.

Alippi, C. and Vanini, G. (2006). A RSSI-based and Cal-

ibrated Centralized Localization Technique for Wire-

less Sensor Networks. In Proceedings - Fourth Annual

IEEE International Conference on Pervasive Comput-

ing and Communications Workshops, PerCom Work-

shops 2006, volume 2006, pages 301–305. IEEE.

Amundson, I., Sallai, J., Koutsoukos, X., and Ledeczi, A.

(2010). Radio interferometric angle of arrival esti-

mation. Lecture Notes in Computer Science, 5970

LNCS:1–16.

Evaluation of Range-based Methods for Localization in Grain Storages

111

ASAE Standards (1999). D245.5 Moisture Relationships of

Plant-based Agricultural Products. ASAE, St. Joseph.

Balakrishnan, H., Baliga, R., Curtis, D., Goraczko, M.,

Miu, A., Priyantha, N., Smith, A., Steele, K., Teller,

S., and Wang, K. (2003). Lessons from develop-

ing and deploying the cricket indoor location system.

Preprint.

Boukerche, A., Oliveira, H., Nakamura, E., and Loureiro,

A. (2007). Localization Systems for Wireless Sensor

Networks. Wireless Communications, IEEE, 14(6):6–

12.

Bulusu, N., Heidemann, J., and Estrin, D. (2000). GPS-

less Low-Cost Outdoor Localization for Very Small

Devices. IEEE Personal Communications, 7(5):28–

34.

Gezici, S. and Poor, H. V. (2009). Position Estimation via

Ultra-Wide-Band Signals. Proceedings of the IEEE,

97(2):386–403.

Hu, L. and Evans, D. (2004). Localization for Mobile Sen-

sor Networks. Proceedings of the 10th annual interna-

tional conference on Mobile computing and network-

ing MobiCom 04, pp(October):45–57.

Juul, J. P., Green, O., and Jacobsen, R. H. (2015). Deploy-

ment of Wireless Sensor Networks in Crop Storages.

Wireless Personal Communications, 81(4):1437–

1454.

Kim, C. W., Chint, F. P. S., and Garg, H. K. (2006). Mul-

tiple Frequencies for Accuracy Improvement in Near

Field Electromagnetic Ranging (NFER). In IEEE In-

ternational Symposium on Personal, Indoor and Mo-

bile Radio Communications, PIMRC, number 3, pages

1–5.

Kumar, S. and Lobiyal, D. K. (2013). An Advanced

DV-Hop Localization Algorithm for Wireless Sen-

sor Networks. Wireless Personal Communications,

71(2):1365–1385.

Langendoen, K. and Reijers, N. (2003). Distributed Local-

ization in Wireless Sensor Networks: A Quantitative

Comparison. Computer Networks, 43(4):499–518.

Levis, P., Madden, S., Polastre, J., Szewczyk, R., White-

house, K., Woo, A., Gay, D., Hill, J., Welsh, M.,

Brewer, E., and Culler, D. (2005). TinyOS: An Op-

erating System for Sensor Networks. In Weber, W.,

Rabaey, J. M., and Aarts, E., editors, Ambient Intel-

ligence, pages 115–148. Springer Berlin Heidelberg,

Berlin, Heidelberg.

Miron, D. B. (2006). Small Antenna Design. In Small An-

tenna Design, pages 56–58. Newnes, 1 edition.

Mühlbauer, W. (1986). Present Status of Solar Crop Drying.

Energy in Agriculture, 5(2):121–137.

Nasipuri, A. and Li, K. (2002). A Directionality Based

Location Discovery Scheme for Wireless Sensor Net-

works. In Proceedings of the 1st ACM international

workshop on Wireless sensor networks and applica-

tions - WSNA ’02, pages 105–111, New York, USA.

ACM Press.

Nelson, S. O. (1965). Dielectric Properties of Grain and

Seed in the 1 to 50-MC Range. Transactions of the

ASAE, 8(1):38–47.

Priyantha, N. B., Chakraborty, A., and Balakrishnan, H.

(2000). The Cricket Location-Support System. In

Proceedings of the 6th annual international confer-

ence on Mobile computing and networking - Mobi-

Com ’00, pages 32–43.

Rappaport, T., Reed, J., and Woerner, B. (1996). Position

Location Using Wireless Communications on High-

ways of the Future. IEEE Communications Magazine,

34(October):33–41.

Rappaport, T. S. (2002). Wireless Communications: Prin-

ciples and Practice. Prentice Hall, 2 edition.

Sathyan, T., Humphrey, D., and Hedley, M. (2011). WASP:

A System and Algorithms for Accurate Radio Local-

ization using low-cost Hardware. IEEE Transactions

on Systems, Man and Cybernetics Part C: Applica-

tions and Reviews, 41(2):211–222.

Schantz, H. (2005). Near Field Phase Behavior. IEEE An-

tennas and Propagation Society International Sympo-

sium, 3B(July):134–137.

Thorbjornsen, B., White, N. M., Brown, A. D., and Reeve,

J. S. (2010). Radio Frequency (RF) Time-of-Flight

Ranging for Wireless Sensor Networks. Measurement

Science and Technology, 21(3):035202.

Whitehouse, K., Karlof, C., and Culler, D. (2007). A Practi-

cal Evaluation of Radio Signal Strength for Ranging-

based Localization. Mobile Computing and Commu-

nications Review, 11(1):41–52.

Xiong, H., Chen, Z., Yang, B., and Ni, R. (2015). TDOA

Localization Algorithm with Compensation of Clock

Offset for Wireless Sensor Networks. China Commu-

nications, 12(10):193–201.

SENSORNETS 2016 - 5th International Conference on Sensor Networks

112