Signal Processing - March 2016 - 66

LOS
Panel

data discussed in [19] and [20] that mimic the intended AL
application scenario. It is shown that a single access point
provides enough position-related information to enable high
accuracy localization. A properly parameterized environment
model is a key ingredient to achieve this.

Right Wall
Right Wall−Cupboard

SINRk [dB]

40
30

Measurement scenario and setup

The mm-wave channel measurements of [19] and [20] are
MIMO measurements with 7 × 7 locations on both transmitter (TX) and receiver (RX) sides obtained by a vector network analyzer. In the intended application, one array assumes
the role of the agent to be localized, while the other corresponds to the anchor, i.e., the fixed infrastructure. The measurement grid on the agent side was moved to 22 different
locations in the room. Both LOS and obstructed LOS (OLOS)
situations have been measured; the latter were obtained using
a laptop screen to shadow the direct link to the anchor. These
measurements have been conducted at a center frequency of
63 GHz. To mimic the IEEE 802.11ad standard [22], we
selected a subband of 2 GHz from the total measured bandwidth of 4 GHz using a raised cosine filter (cf. [33]).

10
0
−10

d3
d4
Training Segment
(a)

0.06
LOS

PEB [m]

0.05

OLOS

NLOS

0.04
0.03
0.02
0.01
0

d3
d4
Training Segment
(b)

Measurement results

We first analyze the SINRs of the MPCs as defined in (2), i.e.,
the ratio of the useful energies of the deterministic MPCs to the
interference by DM and additive white Gaussian noise
figure 6. (a) Estimated SINRs of selected reflections using Tp = 0.5 ns and
(AWGN). The SINRs are estimated using the technique of [18]
fc = 63 GHz and (b) PEB for LOS, OLOS, and NLOS (complete lack of the
and [33], a method of moments estimator operating directly on
LOS component) scenarios. Solid lines indicate LOS measurements; while
dashed and dash-dotted lines correspond to OLOS and NLOS measurements,
the MPC amplitudes. In this way, the PDP S o (x) does not
respectively. The x-axis labeling refers to the measurement sets acquired at
explicitly have to be estimated. We use the array positions on
different positions d 1, d 2, f, d 5 as reported in [19] and [20].
the anchor side to provide the required signal ensemble. The
array at the agent side is used to show the potential of beamforming. In a practical setup, it may be advantageous to impleprice to pay is a very large signal bandwidth to enable the
ment the beamforming at the anchor side, i.e., at the
separation of MPCs at sufficiently high SINRs. Microwaveinfrastructure, where the array has a known orientation, while
band UWB systems can fulfill this promise, but their massat the agent side, low-complexity terminals may be used that
market adoption seems unlikely [9] given the recent
have only one or a few fixed antennas. We reverse these roles
developments of indoor wireless systems. For conventional
here, since the horizontal array geometry at
wireless systems, it would also be possible
the agent side was better suited for a proof
to utilize the phase evolution of the MPCs
The reduction of the
of principle.
for precise localization and tracking [45].
required infrastructure
The estimated SINRs in Figure 6(a)
This technique, however, requires large
show
the relevance of selected MPCs in
arrays for separating the MPCs at moderis of key importance for
this
environment
for several agent positions.
ate bandwidths and might not be relevant
a viable localization
The LOS is the MPC providing most posiin an AL context. On the other hand, 5G
system for AL.
tion-related information. Besides the fact
wireless systems will include UWB radios
that it is usually the strongest component of
in the mm-wave frequency band. The
a radio channel, this significance is due to the relatively low
IEEE 802.11ad standard [22], for example, already defines
impact of DM on the LOS component at a bandwidth of 2 GHz
an air-interface for a 2-GHz bandwidth system in the
[33]. Interestingly, in some cases, the SINR of the LOS com60-GHz frequency band. Beamforming and tracking of
ponent drops only slightly in the OLOS situation, although its
MPCs are key elements of such systems. Despite the promisenergy drops significantly (the average LOS K-factor over the
ing features of mm-wave systems for localization, only few
estimation positions decreases from 8.9 dB to - 7.4 dB). This
papers address this aspect so far, and even fewer discuss
measurement data and realistic channel models [46], [47].
implies that the component is still exploitable for localization.
This section highlights the great potential of mm-wave
The reflected components also show significant SINRs over
technologies for realizing multipath-assisted indoor localizathe estimation points but there is a considerable amount of
tion. We analyze, for this purpose, exemplary measurement
location-dependence of the SINRs. It is more pronounced than
66

IEEE SIgnal ProcESSIng MagazInE

March 2016

Table of Contents for the Digital Edition of Signal Processing - March 2016