[ieee 2009 ieee mtt-s international microwave workshop on wireless sensing, local positioning, and...
TRANSCRIPT
Positioning with moving IEEE 802.15.4 (ZigBee) transponders
M. Pichler1, S. Schwarzer2, A. Stelzer3, and M. Vossiek4
1Linz Center of Mechatronics GmbH, 4040 Linz, Austria2Siemens Corporate Technology, 81739 Munich, Germany
3Johannes Kepler University, Inst. for Communications and Information Engineering, 4040 Linz, Austria4Clausthal University, Inst. of Electrical Information Technology, 38678 Clausthal-Zellerfeld, Germany
Abstract— In an ever-increasing number of wireless communi-cations applications the measurement of positions of transmittersdistributed in space is desired. Demands on low cost and powerconsumption make solutions that allow positioning with existingcommunications hardware during the process of data transmis-sion particularly interesting. In this paper we discuss a methodfor using IEEE 802.15.4 (ZigBee) transmitter nodes with aspecial frequency-hopping scheme for this purpose and show theinsensitivity of this method to transmitter motion mathematicallyas well as by exemplary simulations and measurements.
Index Terms— ZigBee, sensor networks, local positioning, fre-quency hopping.
I. INTRODUCTION
Low-power off-the-shelf communications solutions that si-
multaneously allow the estimation of transmitter positions
provide an additional benefit in applications where low cost
and low maintenance are required. Recently systems and
architectures based on the low-power PHY-layer defined in
IEEE 802.15.4 have been published [1], [2]. Particularly in
indoor environments, however, larger signal bandwidths than
the 5 MHz available in a single IEEE 802.15.4 channel are
required for precise position estimates.
In [3], [4] the authors therefore presented a method for
utilizing the full 80-MHz-bandwidth available in the 2.4-
GHz-ISM-band by employing a particular frequency-hopping
scheme and coherently combining multiple measurements
from different channels to obtain a precise distance estimate.
By this scheme many disturbing effects such as oscillator
errors in both transmitters and receivers are eliminated, and it
is shown how a position estimate for a stationary transponder
T1 can be derived from the slope of the phase of a matched-
filter output as a function of the channel. Due to space
limitations the full signal model cannot be presented in this
context, and the reader is referred to [4].
This paper extends the original model to transmitters T1
that are in motion during the positioning task and shows that
transmitter velocity is, similarly to oscillator errors, to a large
part canceled from the position estimate. Section II will briefly
introduce the system setup and highlight the differences in the
signal model as compared to [4] for moving transmitters. In
Section III simulations and measurements are presented, and
a conclusion is given in Section IV.
Fig. 1. Measurement setup with K receivers R1, . . . , RK , referencetransmitter T2 and transmitter with unknown location T1 moving with
velocity v shaded gray. Superscripts ·[T,R] indicate transmitter T andreceiver R.
II. SETUP AND SIGNALS
The proposed positioning scheme is based on a time-
difference of arrival (TDOA) measurement as it is also de-
scribed e.g. in [5]. As shown in Fig. 1, this scheme employs
a number of receiving base stations R1, . . . , RK distributed
around the region of measurement. For a 1-D measurement
at least two receivers are required, for higher dimensionality
this number increases. A stationary reference transponder T2
and a moving transponder to be located T1 ideally have a
line of sight connection to all base stations. All receiver and
transmitter positions with the exception of that of T1 are
assumed to be known.
Each transmitter consists of a standard IEEE 802.15.4 radio
chip such as the Texas Instruments CC2420 that is clocked
by an external 16-MHz quartz, a microcontroller, and a chip
antenna. Receivers contain an antenna, a downconverter that
mixes the reception in the ISM-band with a 2.4 GHz local
oscillator signal to baseband, a high-speed analog-to-digital
converter, and a signal processing unit. In the current setup
the full ISM bandwidth of 80 MHz is sampled, and channel
separation is performed digitally.
For positioning, each of the two transmitters sends out data
packets in different channels that are received in all base
stations after a time of travel τ [T,R] = r[T,R]
c0proportional
by the speed of light c0 to the distance r between sender Tand receiver R. To utilize the full available bandwidth, the
positioning result is obtained from a sequence of consecutive
IEEE MTT-S International Microwave Workshop on Wireless Sensing, Local Positioning, and RFID (IMWS 2009 - Croatia)
978-1-4244-5062-6/09/$26.00 ©2009 IEEE
Fig. 2. Possible frequency-hopping sequence of channels c[T ]p used
for transmission of packet p by two transmitters T1 and T2.
measurements, for each of which the receptions are correlated
with an ideal data signal that is synthetically generated in the
receivers. By using a particular hopping sequence, a possibility
for which is depicted in Fig. 2, it is ensured that most
perturbing factors such as oscillator frequency error and—as
we will show—transmitter motion can be eliminated during
the computation. The hopping sequence is distinguished by
its symmetry and by a constant packet spacing, both of which
are necessary for the positioning method to work properly.
We assume a model of constant transmitter velocity v during
the short time span of the measurement sequence. In the
absence of noise, the received signal sR is a replica of the
transmitted signal sT that is delayed by the time-of-flight delay
τ[T,R]0 =
r[T,R]0
c0at the beginning of the transmission of the
first data packet, and stretched by the Doppler shift caused by
τ̇0[T,R] =
v[T,R]0
c0, the temporal derivative of τ0 proportional to
the relative radial velocity v0 of the transmitter as seen from
the receiver, i.e.
s[T,R]R (t) = s
[T ]T
(
t −(
τ[T,R]0 + τ̇
[T,R]0 t
))
. (1)
After analog downconversion, sampling, channel separation,
digital conversion to baseband, and decimation, a signal of
the form
s[T,R]F,p [m] = s0,d[T ]
p
(
T[T,R]F
(
m − n[T,R]F,p
)
)
ei(
k[T,R]F,p
m 2π
M+ϕ
[T,R]F,p
)
(2)
with parameters
T[T,R]F = 2π
ωSI
(
1 − τ̇[T,R]0
)
NM
(3)
n[T,R]F,p = ωSI
2π
1
1−τ̇[T,R]0
MN
(
τ[T,R]0 −
ϕ[T ]C0−2πn
[T ]D,p
ωCI− ϕD0
ωDI
)
+ MN
ϕ[R]S0
2π− pM
(4)
k[T,R]F,p = −
ωRI,c[T ]p
ωSINτ̇
[T,R]0 (5)
ϕ[T,R]F,p = ωRI,c[T ]
p
(
− τ[T,R]0 +
ϕ[T ]C0−2πn
[T ]R,p
ωCI−
(
1−τ̇[T,R]0
)
ϕ[R]S0
ωSI
)
+ ϕR0,c
[T ]p
− ϕ[R]L0 + ωLI
ϕ[R]S0
ωSI− ωRI,c[T ]
p
2πτ̇[T,R]0
ωSIpN
(6)
is obtained [4]. In (2)-(6), p denotes the index of the data
packet, s0,dpthe signal transmitted for data packet dp, N
is the number of samples per packet before, M after digital
decimation, ωCI , ωRI,cp, ωLI , and ωSI are the angular trans-
mitter clock frequency, RF center frequency in channel cp,
receiver local oscillator frequency, and sampling frequency,
respectively, ϕC0 is the transmitter main clock phase relative
to time zero, ϕD0 the data signal phase relative to the main
transmitter clock edge number nD,p, ϕL0 and ϕS0 are the
initial phases of the local oscillator and the receiver sampling
clock, respectively, and ϕR0,cpis the transmitter RF signal
phase in channel cp relative to the main transmitter clock
edge number nR,p. The parameters (3)-(6) are computed from
every measurement with respect to an ideal synthetic signal
by correlation.
For proper operation we must demand that the relative offset
between the data clock and the RF signal phase must be
constant for every channel, that is, the offset may be described
by a number of offset clock cycles ∆nR,cponly dependent
on the channel. Furthermore, we have already stated that the
packet spacing in time must be constant after an initial number
of clock cycles nD,0. Hence,
n[T ]R,p = n
[T ]D,p + ∆nR,c[T ]
pn
[T ]D,p = n
[T ]D,0 +
ωCI
ωSI
Np (7)
The employed symmetric frequency-hopping scheme may be
formulated as
c[T ]p = c
[T ]P−1−p c = c[T1]
p1= c[T2]
p2(p1 6= p2), (8)
that is, each transmitter sends a packet in the same channel at
an equal spacing before and after the center of the measure-
ment sequence that is indicated by a dash-dot line in Fig. 2,
and both transmitters occupy each channel exactly twice in
different time slots.
For the following we assume a setup with two receivers R1
and R2 with which a one-dimensional position estimate can
be computed. The measurement can be extended to higher
dimensions e.g. by using all possible receiver pairs for 1D-
estimation and deriving the higher dimension estimate from
multiple 1D results.
From the 8 phase estimates ϕF,p according to (6) that are
obtained in every channel c at packet positions p we can
compute a phase measure ϕF,c that is a function of channel
index c as
ϕF,c =(
ϕ[T1,R1]F,p1
+ ϕ[T1,R1]F,P−1−p1
)
−(
ϕ[T1,R2]F,p1
+ ϕ[T1,R2]F,P−1−p1
)
−(
ϕ[T2,R1]F,p2
+ ϕ[T2,R1]F,P−1−p2
)
+(
ϕ[T2,R2]F,p2
+ ϕ[T2,R2]F,P−1−p2
)
.
(9)
IEEE MTT-S International Microwave Workshop on Wireless Sensing, Local Positioning, and RFID (IMWS 2009 - Croatia)
978-1-4244-5062-6/09/$26.00 ©2009 IEEE
After insertion of (6) into (9) we obtain ϕF,c as
ϕF,c ≅ 2ωRI,c
(
− τ̃ [T1,R1] + τ̃ [T2,R1] + τ̃ [T1,R2] − τ̃ [T2,R2]
+
(
τ̇[T1,R1]0 − τ̇
[T2,R1]0
)
ϕ[R1]S0 −
(
τ̇[T1,R2]0 + τ̇
[T2,R2]0
)
ϕ[R2]S0
ωSI
)
(10)
where
τ̃ [T,R] = τ[T,R]0 + τ̇
[T,R]0
2π
ωSI
(P − 1)N
2. (11)
With an expression for a phase ϕF,c as a function of channel
index c we have now obtained the quantity of interest. As
ϕF,c depends on the channel center frequency ωRI,c, it is
a linear function with a slope ideally proportional to the
desired TDOA −τ̃ [T1,R1] + τ̃ [T2,R1] + τ̃ [T1,R2] − τ̃ [T2,R2].
For a stationary reference transponder T2, τ̇[T2,Rk]0 = 0 and
τ̃ [T2,Rk] = τ[T2,Rk]0 ∀k, which further simplifies the result.
The remaining terms can be split in two parts. The first line
in (10) describes the TDOA corresponding to the position of
the transmitter at the center of the frequency-hopping sequence
that is indicated by a dash-dot line in Fig. 2. From the phase
terms ϕF,c it can be estimated e.g. as the slope of a linear
least squares fit or by a Fourier-analysis [3], [4]. The second
line contains the error terms remaining even after application
of the frequency-hopping scheme. These are dependent on
the initial phase of the receiver sampling clocks, that is,
on the synchronization mismatch between the receivers. For
equal starting points of the sampling process in the receivers
all ϕ[Rk]S0 are identical and this error vanishes. Otherwise, a
small error in the phase slope remains, but it is shown to
be negligible for realistic system parameters in the following
section. All other unknown perturbing terms such as initial
oscillator phases and, most importantly, terms depending on
transponder velocity have been eliminated in (10).
III. SIMULATION AND MEASUREMENT
For verification of the claimed insensitivity to transmitter
motion simulations in 1D- and 2D-scenarios and measure-
ments in a 2D-scenario have been carried out.
Firstly, a simulation of the effect of the phase slope error
caused by transmitter velocity and unsynchronized receivers
was performed. Basis of the simulation is a 1-dimensional
setup with receivers R1 at 0 m and R2 at 10 m, respectively,
the reference transmitter T2 at 5 m, and the transmitter T1 to be
located at 2.5 m. For demonstration the moving transmitter was
assumed to have a positive velocity of τ̇[T1,R1]0 c0 = 50 m/s,
which is high for all relevant applications. According to (10)
the influence of the receiver synchronization mismatch on the
position estimate is equal to the negative transmitter velocity,
that is,d δr
d(ϕ
[R2]
S0
ωSI
)
= −50m/s. (12)
ϕ[R2]
S0
ωSI(µs)
δr(µ
m)
−5 −4 −3 −2 −1 0 1 2 3 4 5−300
−200
−100
0
100
200
300
Fig. 3. Simulation of the range estimation error δr for a transmittermoving at 50 m/s and a receiver synchronization error between −5
and 5 µs.
Fig. 3 shows the simulation result that is aside from numerical
error in the correlation in accordance with the theoretical value
(12), so the line exhibits a slope of −50 µm/µs. Note that even
for a high velocity and receiver synchronization imperfection
of 5 µs the distance error of only 250 µm is negligible.
Secondly, two-dimensional simulations and measurements
were performed. The setups can be seen in Fig. 4 and 5. They
consisted of four receivers R1, . . . , R4 set up in a rough square
around the region of interest. The reference transponder T2
was placed in the center and the measurement transponder T1
was moving diagonally across the field of measurement and on
a circle around the center in the two scenarios. The estimate of
the 2-dimensional transponder position was obtained from the
measurements of each of 6 possible pairs of receivers using an
optimization criterion over the xy-plane from which estimates
in x- and y-directions were derived.
In the simulation the velocity of the transponder was set to
a high value of 50 m/s for the motion on the diagonal, and
the angular velocity was set to 19.6 rad/s for the motion on a
circle with 2.5 m radius, which corresponds to a translational
velocity of 49 m/s. In Fig. 4 the estimation results for the
position of T1 are shown, where gray dots mark transmitter
positions at the beginning of the measurement, gray lines
the position estimate projection to the temporal center of
the frequency hopping sequence due to transmitter velocity,
and small black dots the actual position estimates that should
ideally be at the tip of the position projection vectors. In both
scenarios there is obviously little error between the theoretical
value and the estimate.
Measurements with a setup similar to that of the simulation
were carried out, with a person carrying the moving transpon-
der T1 on a diagonal and a circular path in two scenarios, as
it can be seen in Fig. 5. While only a qualitative conclusion
can be drawn from this measurement result, it demonstrates
that the proposed IEEE 802.15.4 positioning method works
not only for stationary, but also for moving transponders.
IEEE MTT-S International Microwave Workshop on Wireless Sensing, Local Positioning, and RFID (IMWS 2009 - Croatia)
978-1-4244-5062-6/09/$26.00 ©2009 IEEE
Fig. 4. Simulation results for 2-dimensional setup consisting of four receivers R1, . . . , R4, stationary reference transmitter T2 andmoving transmitter T1 with position projection vectors. Black dots indicate position estimates.
Fig. 5. Measurement results for 2-dimensional setup consisting of four receivers R1, . . . , R4, stationary reference transmitter T2 andmoving transmitter T1. Black dots indicate position estimates.
IV. CONCLUSION
In this paper an extension of a proposed localization
method using IEEE 802.15.4 (ZigBee) off-the-shelf transmitter
chips and data signals for positioning to the case of moving
transponders was presented. It was shown that most error
terms due to the Doppler shift that are present in received
signals are canceled by a proposed symmetric frequency-
hopping sequence. Via a simulation it was verified that the
remaining position estimation error that depends on receiver
synchronization imperfection will be negligible for reasonable
parameters. The method has been shown to work also for fast-
moving transponders in 1- and 2-dimensional scenarios.
ACKNOWLEDGEMENT
This work was sponsored by the “Austrian Center of
Competence in Mechatronics (ACCM)” within the COMET
Program of the Austrian federal government, the Province of
Upper Austria, and the Scientific Partners of ACCM.
REFERENCES
[1] P. Havinga, B. Dil, and M. Bijl, “Localization schemes integration withzigbee communication protocol stack,” presented at 1st European ZigBee
Developer’s Conference, Munich, Germany, June 18-20, 2007.[2] S. Lanzisera, D. Lin, and K. Pister, “RF time of flight ranging for wireless
sensor network localization,” in Proc. 2006 International Workshop
on Intelligent Solutions in Embedded Systems (WISWS 2006), Vienna,Austria, June 30, 2006, pp. 1–12.
[3] S. Schwarzer, M. Vossiek, M. Pichler, , and A. Stelzer, “Precise distancemeasurement with IEEE 802.15.4 (ZigBee) devices,” in Proc. IEEE Radio
and Wireless Symposium (RWS 2008), Orlando, FL, USA, Jan. 22-24,2008, pp. 779–782.
[4] M. Pichler, S. Schwarzer, A. Stelzer, and M. Vossiek, “Multi-channel dis-tance measurement with IEEE 802.15.4 (ZigBee) devices,” IEEE Journal
of Selected Topics in Signal Processing, Oct. 2009, to be published.[5] A. Stelzer, K. Pourvoyeur, and A. Fischer, “Context and application of
LPM—A novel 3-D local position measurement system,” IEEE Trans.
Microwave Theory Tech., vol. 52, no. 12, pp. 2664–2669, Dec. 2004.
IEEE MTT-S International Microwave Workshop on Wireless Sensing, Local Positioning, and RFID (IMWS 2009 - Croatia)
978-1-4244-5062-6/09/$26.00 ©2009 IEEE