On Clustering for Deterministic and Measured Indoor mmW Channels

Maria-Teresa Martinez-Ingles1, Davy P. Gaillot2, Juan Pascual-Garcia2, Jose-Maria Molina Garcia-Pardo1,Martine Lienard2, José-Víctor Rodriguez1

1 Universidad Politécnica de Cartagena, Dpto. Tecnologías de la Información y las Comunicaciones, Cartagena, Murcia, Spain. e-mails: {mteresa.martinez, josemaria.molina, juan.pascual, jvictor.rodriguez}@upct.es

2 University of Lille 1, IEMN/TELICE, Villeneuve d’Ascq, France e-mails: {davy.gaillot, martine.lienard}@univ-lille1.fr

Abstract—This work presents the multidimensional analysis

of both deterministic and measurements data in terms of clusters for indoor mmW channels. Firstly, by applying RiMAX, MPC (multipath components) and DMC (dense multipath components) are extracted from both an indoor measurement campaign and ray tracing simulations, including diffuse scatter. Secondly, from the RiMAX estimated paths, K-means with MCD (multipath component distance) is applied to compare the distribution of clusters.

Index Terms—mmW, ray tracing, Rimax, MCD, cluster.

I. INTRODUCTION The mmW frequency-band has emerged as a very

promising response to future high-data rate wireless communication systems [1]. Prediction of the physical layer is essential to study the performance of mmW systems. Ray-tracing has shown to be an accurate tool to study propagation at those frequencies, considering all possible propagation mechanisms, i.e. line of sight, reflection, diffraction and diffuse scattering. On the other hand, measurement campaigns permit the extraction of important statistical parameters, which can feed propagation models to improve their accuracy.

In this paper, we analyze the behavior of clusters in an indoor environment, by means of a measurement campaign and simulations with a ray-tracing tool. The main contribution of this work is to apply the RiMAX estimator, recently developed by the authors, on both simulated and measured frequency channel transfer functions. In the case of simulations, we have also considered diffuse scattering [2]. Then, the clusters are identified using the K-means and MCD (Multipath Component Distance). This approach aims to verify whether this deterministic tool can accurately predict the main inter-clusters parameters at mmW frequencies.

II. SCENARIO, MEASUREMENTS AND SIMULATIONS In this section, we will briefly present the investigated

scenario, the measurement setup, and protocol as well as the ray-tracing tool.

A. Scenario The scenario is a laboratory of the Universidad Politécnica

de Cartagena. This laboratory consists of a room of dimensions 4.5 × 6.5 × 2.5 meters and it is furnished with several closets,

desktops, and computers. The walls are made of plasterboard, whereas the floor is made of concrete. More information about this matter can be found in [3].

B. Measurements The measurements were conducted using a channel sounder

based on the Rohde ZVA67 Vector Network Analyzer (VNA). The measured frequency range was 57–66 GHz using 4096 frequency points. A 10 Hz intermediate frequency was selected and a dynamic range of more than 100 dB was obtained. Two amplifiers are used in the transmission to compensate for the attenuation of the cables (HXI HLNA-465). The system is THOROUGH calibrated to eliminate the effect of cables and amplifiers. The transmission power of the VNA was set to 22 dBm.

Both Tx and Rx antennas are omnidirectional antennas (Q-par QOM55-65 VRA) with 4.5 dBi gain, with their polarization being always vertical. The height of the transmitting and receiving antenna was 1.44 m. The receiving antenna is moved along a linear positioner within 5 positions separated by 2 mm (ULA), whereas the transmitting antenna is moved using a 6 x 6 uniform rectangular array (URA), also by 2 mm steps along both dimensions.

C. Simulations The 3D ray-tracing technique employed in this work is

fully written in Matlab. As usually, reflected and diffracted components are computed. Components produced by the reflection of a diffracted wave and components produced by the diffraction of a reflected wave are also considered. Furthermore, diffuse components have been simulated in order to increase the accuracy of the model [2]. The maximum number of reflections is two, whereas the maximum number of reflections after diffraction is one. These values assured a proper convergence in all simulations. No substantial differences were found between two and three reflections.

III. RIMAX AND CLUSTERING

A. RiMAX The measured MIMO radio channels were processed with

the RiMAX maximum-likelihood algorithm which extracts from the sampled radio channels the propagation paths parameters (time-delays, azimuth angles and complex gains)

The 8th European Conference on Antennas and Propagation (EuCAP 2014)

978-88-907018-4-9/14/$31.00 ©2014 IEEE 802

and Dense Multipath Components (DMC) [4]. The DMC is stochastic by nature and includes both diffuse scattering and all paths that cannot be resolved. The incorporation of the DMC components into the data model was shown to improve the accuracy and validity of the estimated propagation paths [5].

512 frequency points were chosen to obtain a 1.12 GHz bandwidth out of the 9 GHz total bandwidth. In addition, the whole MIMO array was selected to perform the estimation (6 x 6 URA for the transmitter and 5-element ULA for the receiver). The ULA restricts the angular estimation between -90° and 90° but is here sufficient to grasp the physics of the propagation mechanisms. Finally, we noted that the data model considers antennas with isotropic radiating properties for all frequencies and, therefore, does not take into account the broadband antenna patterns.

B. Clustering For clustering the data, we use the K-means algorithm

which iteratively moves a number of cluster centroids through the data space to minimize the total difference between MPCs and their closest centroid. In addition, we have used the Multipath Component Distance (MCD) [6-7].

IV. RESULTS Figure 1 presents the AoA (Angle of Arrival)/AoD (Angle

of Departure) MPC results for a) the simulated data, and b) the measured data for one position in the measured room. MPCs are grouped with either color circles or crosses and the 5 clusters with most energy are represented by circles (Fig. 1).

From these results, the main intra-cluster parameters of the first 5 clusters were obtained [8]. Table I presents the power, rms DS (delay spread), AoA and AoD spreads for each cluster. Note that, since most of the information is included in the azimuth plane, the elevation results were skipped but it will be discussed in the full paper.

V. CONCLUSIONS In this work, we have studied the multidimensional

properties of both deterministic models and measurements in terms of clusters for indoor mmW channels. Firstly, RiMAX was applied to both data. Secondly, K-means with MCD were applied from the extracted MPCs to identify the cluster distribution. A relative good agreement is found between the

simulated and measured results indicating that the Ray Tracing tool is a valid approach to determine intra-clusters parameters.

TABLE I. PARAMETERS OF FIRST FIVE CLUSTERS

Simulations / Measurements Cluster Power (dB) DS (ns) AoAS (º) AoDS (º)

1 0/0 1.63/2.7 14.7/13.2 13.27/23.14

2 -10.2/-10.1 2.04/2.5 11.32/10.6 10.34/12.26

3 -10.4/-11.1 4.5/15 20.29/3.23 35.47/28.84

4 -11.4/-12.4 2.5/9.42 9.02/3.29 20.81/11.28

5 -13.4/-12.8 4.27/6.06 8.02/9.12 11.75/10.37

VI. ACKNOWLEDGMENT This work was supported by MINECO, Spain (TEC2010-

20841-C04-03) and by the European FEDER funds, and by COST IC-1004 by STSM.

REFERENCES [1] P. Smulders, "Exploiting the 60 GHz band for local wireless multimedia

access: prospects and future directions," Communications Magazine, IEEE, vol.40, no.1, pp.140-147, Jan 2002.

[2] V. Degli-Esposti, F. Fuschini, E. M. Vitucci, and G. Falciasecca, Measurement and Modelling of Scattering From Buildings, Transactions on Antennas and Propagation, vol.. 55, no. 1, January 2007.

[3] M.T. Martinez-Ingles, J. Pascual-Garcia, J.V. Rodriguez, J.M. Molina-Garcia-Pardo, L.J. Juan-Llacer, D. P Gaillot, M. Lienard and P. Degauque, , “Initial experimental characterization of the millimeter-wave radio channel”, 7th European Conference on Antennas and Propagation (EUCAP), pp. 2796-1799, 8-12 April 2013.

[4] A. Richter, “Estimation of radio channel parameters: models and algorithms”, Thesis, TU Ilmenau, 2005.

[5] E. Tanghe, D. P. Gaillot, W. Joseph, M. Liénard, P. Degauque, and L. Martens, “Robustness of high-resolution channel parameter estimators in the presence of dense multipath components,” IET Electronics Letters, vol. 48, no. 2, pp. 130–132, January 2012.

[6] Steinbauer, M., et al.: ‘How to quantify multipath separation’, IEICE Trans. Electron., 2002, E85, (3), pp. 552–557.

[7] Nicolai Czink, Pierluigi Cera, Jari Salo, Ernst Bonek, Jukka-Pekka Nuutinen, Juha Ylital, A framework for automatic clustering of parametric MIMO channel data including path powers, Vehicular Technology Conference, 2006. VTC-2006 Fall. 2006 IEEE 64th.

[8] J. Salmi, A. Richter, and V. Koivunen, “Detection and tracking of MIMO propagation path parameters using state-space approach,” IEEE Transactions on Signal Processing, vol. 57, no. 4, pp. 1538–1550, April, 2009.

-100 -80 -60 -40 -20 0 20 40 60 80 100-200

-150

-100

-50

0

50

100

150

200

250

AoA azimut (º)

AoD

azi

mut

(º)

-100 -80 -60 -40 -20 0 20 40 60 80 100

-200

-150

-100

-50

0

50

100

150

200

250

AoA azimut (º)

AoD

azi

mut

(º)

Figure 1: Extracted MPC using RiMAX from a) Simulations and b) Measurements

Cluster 1

Cluster 2

Cluster 3

Cluster 4

Cluster 5

The 8th European Conference on Antennas and Propagation (EuCAP 2014)

803