7th IEEE International Symposium on Industrial Embedded Systems (SIES'12)

Performance Evaluation of Chirp Spread Spectrum

Ranging for Indoor Embedded Navigation Systems
Paolo Pivato∗ , Stefano Dalpez† and David Macii∗
∗ Department of Information Engineering and Computer Science (DISI)
University of Trento, Via Sommarive, 5 - 38123, Trento - Italy
E-mail:{pivato,macii}@disi.unitn.it
†
Centre for Materials and Microsystems (CMM IRST)
Fondazione Bruno Kessler, Via Alla Cascata, 56 - 38123, Trento - Italy
E-mail: [email protected]

Abstract—Wireless distance measurement techniques based on and the message timestamping mechanism. The main features
portable embedded platforms are expected to play a key role in of some wireless ranging solutions are shortly summarized in
several industrial and domestic applications. In this paper the Tab. I. A precise comparison between them is difficult, mainly
ranging accuracy of a commercial Chirp Spread Spectrum (CSS)
kit is evaluated experimentally in a real-world context. The pro- because the experimental conditions are not the same and
posed analysis provides more precise and exhaustive information sometimes they are not even clearly described in the respective
than what it is usually reported in the technical literature. In fact, papers. At the moment, the ultra-wideband (UWB) techniques
this paper is specifically focused on performance evaluation and assure the best accuracy (e.g., in the order of 10-20 cm),
it deals with the case of short-range indoor scenarios in both albeit at the expense of a significant power dissipation for a
Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) repeatable
conditions. The resulting analysis represents the first step towards given range [5]–[7]. Measuring the RTT of IEEE 802.15.4
the design of a custom indoor embedded navigation system for message can provide good ranging accuracy in static LOS
a smart rollator assisting impaired people to move safely in an conditions and after an adequate calibration [8]. However, the
indoor public environment. distance measurement error in dynamic scenarios may easily
Index Terms—Indoor Ranging, Chirp Spread Spectrum (CSS), exceed 1 m, unless sophisticated data processing algorithms
Ambient Assisted Living, Embedded Navigation Systems. are used [9]. Alternative solutions based on chirp spread
spectrum (CSS) exhibit promising accuracy (i.e., below 50 cm)
I. I NTRODUCTION especially outdoors [10]–[12]. However, the existing data on
In the last few years a growing body of research has CSS-based ranging performance in indoor environments are
faced the issue of indoor localization. Several alternative confusing. For this reason, this paper reports the results of
solutions for indoor ranging and localization have been pro- a thorough empirical characterization of a commercial CSS-
posed, e.g. based on laser rangefinders, ultra-sound devices, based ranging platform under clear and repeatable experimen-
infrared sensors, radio frequency (RF) signals and video cam- tal conditions. The proposed analysis represents the first step
eras [1]–[3]. Among them, RF ranging techniques are ideally towards the design of a novel indoor navigation system. Such
preferable because portable embedded systems are already a system is going to be installed on a collaborative rollator,
natively equipped with wireless chips for radio communica- called c-walker, assisting impaired users to move safely within
tion. Unfortunately, distance estimation based on received- a public environment such as a shopping mall or an airport.
signal-strength (RSS) measurements is strongly affected by
multipath propagation phenomena, especially indoors. In fact, II. CSS- BASED R ANGING
the received signal intensity generally exhibits large position-
dependent fluctuations, instead of a monotonic behavior, even CSS is a spread spectrum radio communication technique
in LOS conditions [4]. This in turn strongly influences ranging relying on linear frequency modulated signals. In particular,
accuracy, which indeed can be hardly smaller than a few CSS systems use chirp signals to code the user data stream.
meters. The signal-to-noise ratio (SNR) at the receiving end (i.e. the
In order to address this problem, some alternative wireless processing gain) is improved by means of a pulse compression.
technologies and methods have been proposed to estimate the When a chirp of duration Tc is processed by its matched filter,
distance between two radio devices, e.g. by measuring the the output peak has a duration proportional to 1/B, where B
time of arrival (TOA), the time difference of arrival (TDOA) is the chirp bandwidth [13]. Hence, the processing gain can
or the round trip time (RTT) of radio messages. When some be easily scaled by varying the bandwidth-duration product
message time-of-flight parameter is measured, the influence of of the chirp. Given an incoming chirp signal s(t, ωi , Ti ) with
multipath propagation on distance estimation can be mitigated center frequency ωi at time Ti , the peak at the output of the
by properly choosing both the radio communication scheme matched filter reaches its maximum at t = Ti . Accordingly,

978-1-4673-2684-1/12/$31.00 ©2012 IEEE 307

 TABLE I
ACCURACY PERFORMANCE OF DIFFERENT TOF- BASED
RANGING SYSTEMS .

Authors Technology Method Accuracy1

Falsi et al. [5] UWB TOA ≈ 15 cm

De Angelis et al. [7] UWB RTT ≈ 20 cm

Tingting et al. [6] UWB TDOA ≈ 20 cm

Santinelli et al. [8] ZigBee RTT ≈ 60 cm

Röhrig et al. [12] CSS RTT ≈ 50 cm

1 - The definition of accuracy may slightly differ in different papers.
The reported values generally include the effect of both mean error
and standard deviation.

the matched filter output signal has a sinc-like shape given by:
µ  Fig. 1. Screenshot of the NI LabView application for data acquisition.
sin 2 ·(t−Ti )·(Tc −|t−Ti |)
g(t,ωi ,Ti )= µ ·ejωi (t−Ti ) |t−Ti |<Tc (1)
2
symmetric double-sided two-way ranging (SDS-TWR) [15].
where µ is the chirp rate expressed in rad/s2 . The amendment
In the experimental sessions presented in this paper, one
IEEE 802.15.4a-2007 recommends the CSS as an alternative
device was kept steadily in a reference position, whereas the
physical (PHY) layer for wireless sensor networks (WSNs)
other was mounted on top of a rollator. In the following the
deployed in high mobility scenarios [14]. It is worth empha-
stationary device will be simply called fixed anchor (FA),
sizing that the duration of CSS-based cross-correlation peaks
while the other one will be referred to as moving target (MT).
can be made much shorter than using the Direct Sequence
In general, the distance values measured by either device
Spread Spectrum (DSSS) signals specified in the standard
are visualized on a dot matrix liquid crystal display (LCD).
IEEE 802.15.4. This property leads to two further benefits.
Since the ranging kit does not offer the possibility to save
First of all, the peak rising time as well as the jitter affecting
a continuous stream of data, a NI LabView application for
message timestamping at the receiving end can be considerably
optical character recognition (OCR) was developed to collect
reduced. This means that, since precise short pulses (e.g.
and to process large amounts of measurement results. The
UWB) are difficult to generate in the time domain, CSS can be
control panel of this application is shown in Fig. 1. The
regarded as an appealing solution for time-of-flight parameter
LabView application at first records the raw images collected
measurement. Secondly, narrow cross-correlation pulses are
by a typical webcam. Each collected image is pre-processed to
generally easier to detect and to discriminate. This is of crucial
increase its contrast and brightness. Afterwards, it is processed
importance because generally possible signal replicas due to
by a set of pattern recognition functions provided by the
multipath propagation may overlap and hide the LOS term.
NI Vision toolbox. These functions, based on a training set of
However, if the peaks of different replicas are distinguishable,
decimal characters, recognize the digits shown on the display.
the probability of detecting the LOS contribution is higher.
Finally, the virtual instrument checks the consistency of the
Therefore, the distance from the transmitter can be estimated
recognized characters by matching the results with a known
with better accuracy.
4-digit pattern (along with the decimal point). In this way,
III. E XPERIMENTS D ESCRIPTION possible invalid values are automatically discarded, whereas
the correct ones are saved into a file in a tabular format.
A. Experimental Setup
Several experiments were conducted in a rectangular room B. Experimental Results
with a size of 8.6 × 6 m. The chosen room has a wooden floor Different kinds of experiments were repeated using the setup
and it is furnished with a table placed at about 1 m from one described in Section III-A. At first, the FA was placed on top
of the side walls. The tool used to evaluate the performance of a 1-m plastic pole at about 1 m from the center of one of the
of CSS-based ranging is the Nanotron Ranging Kit RK-II. longer walls of the room. Antennas were kept perfectly parallel
It consists of two identical battery-operated wireless hand- to one another so as to maximize the communication gain. The
held devices. Each device is equipped with an omnidirectional distance values were collected in LOS conditions with the MT
monopole antenna and a CSS nanoPAN 5375 RF Module located in ten equally spaced points from the FA in the range
transmitting in the 2.4 GHz Instrumentation Scientific and [0, 5] m along three different directions (represented by solid
Medical (ISM) band. The nanoPAN 5375 RF Module is able to lines in Fig. 2). In any of such points the height of the MT
estimate the distance from another twin module through RTT from the floor was varied between 75 cm and 125 cm. On the
measurements. This technique sometimes is also referred to as whole, 4500 data were collected. Fig. 3 shows the box-and-

308
 5

Error [m]
0

−1

−2

−3

−4

−5
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Fig. 2. Experimental setup scheme. Real Distance [m]

whiskers diagram of the measured distance errors in any of Fig. 3. Box-and-whisker diagram of the measured distance errors in static
such points when the MT was 100 cm above the floor. The conditions. The nodes measuring the range are placed at fixed distances from
results for 75 cm and 125 cm are quite similar and are not one another between 0 and 5 m. In each position multiple data are collected.
shown for the sake of brevity. This means that a small offset
between the equatorial planes of the antennas does not affect
5
accuracy significantly. Due to some residual multipath-related,
4
position-dependent offsets, the maximum error is in the order
of ±1 m with a very small standard deviation, i.e. in the order 3

of less than 10 cm. It is worth noting that the data collected 2

at distances shorter than 1 m could be affected by some near- 1
field antenna anisotropies.
Error [m]

0
In order to analyze the performance of the ranging technique
−1
in dynamic conditions, two orthogonal kinds of experiments
were conducted. In both cases the MT was mounted on top −2

of the rollator at about 1 m above the floor. In the first −3

group of experiments a user repeatedly pushed the rollator −4
forth and back along five arcs of a circle with radius equal
−5
to 1 m, 2 m, 3 m, 4 m and 5 m, respectively. In all cases 1 2 3 4 5
Real Distance [m]
the FA was approximately located in the center of the circle
always in LOS conditions. In order to make the intrinsic
uncertainty associated with the movement over the chosen
Fig. 4. Box-and-whisker diagram of the measured distance errors in dynamic
trajectory as little as possible (i.e. in the order of a few cm), conditions with the MT mounted on a rollator moving forth and back over
some adhesive markers were put on the floor to guide the user. five arcs of a circle with radius equal to 1 m, 2 m, 3 m, 4 m and 5 m.
The corresponding measurement errors at different nominal
distances are shown in the box-and-whiskers diagram of Fig. 4.
Observe that, unlike the static case, the mean values of the Fig. 5. In this case the rollator was pushed forth and back
distance error is around zero, but the standard deviation is repeatedly along the same three straight line trajectories
larger than before (i.e., about 30 cm). However, except for a previously considered in the static experiments. The values
few outliers (highlighted by cross markers), the peak-to-peak measured by the MT were collected in 15 equally spaced
error ranges are comparable. Thus, the apparent differences points from the FA in either direction. The box-and-whiskers
between Figs. 3 and 4 are probably due to the fact that diagram in Fig. 5(a) just refers to 30 movements from 5 m
the residual position-dependent systematic offsets change even to 0 m. The diagram shown in Fig. 5(b) instead shows the
when the distance is the same. In other words, the variance of error ranges obtained when the rollator was moving in the
the error distribution in Fig. 4 is larger because it includes opposite direction. No significant differences were observed
the offset variations as a function of time associated with between the results associated with the three chosen trajec-
the movement of the target. Note also that the increment tories. Observe that both the mean value and the standard
of the worst-case absolute error with distance is negligible. deviation of the distance errors in Fig. 5(a) are very different
This means that the relative measurement uncertainty tends to from those in Fig. 5(b). In the former case the two handheld
become smaller when the range becomes larger. devices of the ranging kit are always in LOS conditions and
The results of the dual set of experiments are shown in the measurement accuracy is compatible to what it is shown

309
 5
Since data sheets and other published results typically re-
port unclear information on CSS-based ranging accuracy in
4
indoor environments, in this paper we analyzed empirically
3 the performance of this technique using a last-generation, off-
2 the-shelf kit. Multiple results show that in LOS conditions
1 the accuracy is generally quite good (i.e., typically in the
Error [m]

0
order of 50 cm). On the contrary, in NLOS conditions or in
presence of obstacles, accuracy and precision may degrade
−1
considerably. Therefore, in view of designing a custom indoor
−2 navigation system, occlusion detection will be essential to keep
−3 the positioning uncertainty below specified limits.
−4 V. ACKNOWLEDGEMENTS
−5
0.8 1.1 1.4 1.7 2 2.3 2.6 2.9 3.2 3.5 3.8 4.1 4.4 4.7 5
The research activities and the results presented in this
Real Distance [m] paper are part of the FP-7 project DALI - Devices for Assisted
(a) Living co-funded by the European Union, grant agreement no.
5
ICT 2011-288917.
4 R EFERENCES
3 [1] E. Aitenbichler and M. Muhlhauser, “An IR local positioning system
for smart items and devices,” in Proc. 23rd Int. Conf. on Distributed
2 Computing Systems Workshops. IEEE, May 2003, pp. 334–339.
[2] L. Angrisani, A. Baccigalupi, and R. Schiano Lo Moriello, “Ultrasonic
1
time-of-flight estimation through unscented Kalman filter,” IEEE Trans.
Error [m]

0 Instrum. Meas., vol. 55, no. 4, pp. 1077–1084, Aug. 2006.

[3] B. Andó, S. Baglio, S. La Malfa, and V. Marletta, “A Sensing Architec-
−1 ture for Mutual User-Environment Awareness Case of Study: A Mobility
−2
Aid for the Visually Impaired,” IEEE Sensors J., vol. 11, no. 3, pp. 634–
640, Mar. 2011.
−3 [4] P. Pivato, L. Palopoli, and D. Petri, “Accuracy of RSS-Based Centroid
Localization Algorithms in an Indoor Environment,” IEEE Trans. In-
−4
strum. Meas., vol. 60, no. 10, pp. 1–10, Oct. 2011.
−5 [5] C. Falsi, D. Dardari, L. Mucchi, and M. Z. Win, “Time of arrival
0.8 1.1 1.4 1.7 2 2.3 2.6 2.9 3.2 3.5 3.8 4.1 4.4 4.7 5 estimation for UWB localizers in realistic environments,” EURASIP J.
Real Distance [m]
Appl. Signal Process., vol. 2006, pp. 152–152, Jan. 2006.
(b) [6] Z. Tingting, Z. Qinyu, Z. Naitong, and X. Hongguang, “Performance
Fig. 5. Box-and-whisker diagrams of the measured distance errors in dynamic Analysis of an Indoor UWB System Based on Energy Detection,” in 7th
conditions with the MT mounted on a rollator moving repeatedly over a Annual Communication Networks and Services Research Conference,
straight line from 5 m to 0 m (a) and backwards (b). 2009, CNSR 09., May 2009, pp. 305–309.
[7] A. De Angelis, M. Dionigi, A. Moschitta, and P. Carbone, “A Low-Cost
Ultra-Wideband Indoor Ranging System,” IEEE Trans. Instrum. Meas.,
in Fig. 4, although the variance is a bit smaller. Conversely, vol. 58, no. 12, pp. 3935–3942, Dec. 2009.
[8] G. Santinelli, R. Giglietti, and A. Moschitta, “Self-calibrating indoor
in Fig. 5(b) (i.e., when the MT is completely covered by the positioning system based on ZigBee devices,” in Proc. IEEE Instrumen-
body of the user pushing the rollator) error patterns may reach tation and Measurement Technology Conference, 2009. I2MTC 09, May
3 m and exhibit significant position-dependent fluctuations. 2009, pp. 1205–1210.
[9] D. Macii, F. Trenti, and P. Pivato, “A Robust Wireless Proximity
This is probably due to the excessive attenuation of the peak Detection Technique based on RSS and ToF Measurements,” in Proc.
associated with the received LOS component, which is no IEEE Int. Work. on Measurements & Networking, M&N2011, Oct. 2011,
longer strong enough to dominate over the peaks of the other pp. 31–36.
[10] A. Sikora and V. F. Groza, “Fields Tests for Ranging and Localization
scattered signal replicas. Such results partially confute the with Time-of-Flight-Measurements Using Chirp Spread Spectrum RF-
widely diffused belief according to which CSS-based ranging devices,” in Proc. IEEE Instrumentation and Measurement Technology
should be more robust to non-LOS propagation. In fact, CSS- Conference, 2007. IMTC 07, May 2007, pp. 1–6.
[11] H.-S. Ahn, H. Hur, and W.-S. Choi, “One-way ranging technique for
based ranging looks more robust to multipath propagation if CSS-based indoor localization,” in 6th IEEE International Conference
the LOS component prevails over the other contributions or on Industrial Informatics, 2008, pp. 1513–1518.
when the time-of-flight measurement jitter is small compared [12] C. Röhrig and M. Muller, “Localization of Sensor Nodes in a Wireless
Sensor Network Using the nanoLOC TRX Transceiver,” in Proc. IEEE
to the actual LOS signal propagation time (as it typically Vehicular Technology Conf., Apr. 2009, pp. 1–5.
occurs outdoors). [13] M. E. Skolnik, Radar Handbook, 3rd Edition. McGraw-Hill, 2008, ch.
8 – Pulse Compression Radar, pp. 8.1–8.29.
IV. C ONCLUSIONS [14] IEEE Wireless Medium Access Control (MAC) and Physical Layer
(PHY) Specifications for Low-Rate Wireless Personal Area Networks
Chirp Spread Spectrum (CSS) modulation is considered (WPANs), IEEE Std. 802.15.4a-2007 (Amendment to IEEE Std 802.15.4-
2006), 2007.
as one of the most promising solutions for both wireless [15] R. Hach, “Symmetric Double Sided Two-Way Ranging,” IEEE P802.15
communication and ranging and it is expected to be used Working Group for Wireless Personal Area Networks (WPAN), Doc.
in various portable embedded systems in the near future. IEEE P.802.15-05-0334-00-004a, Jun. 2005.

310