TIME VARYING PATH SHADOWING MODEL FOR INDOOR POPULATED ENVIRONMENTS

Submitted By: Sneha Rachel Philip S1 MTech Roll No. 16

Seminar Guide: Mohammed Salih K. K.

ABSTRACT
This paper presents a path-shadowing model in indoor populated environments based on computer simulations. The propagation paths between the transmitting and receiving points in an empty rectangular space are determined by a ray tracing method, moving quasi-human bodies modelled as a cylinder with finite height are generated in this space, and intersections of the paths with the bodies are counted. As a result, the shadowing probabilities, durations, and intervals are evaluated for each propagation path, and this shadowing process is characterized as a Markov process. This paper proposes a method for individually generating the shadowing effects on each propagation path. The results of 5.2 GHz propagation measurements in indoor populated space using a high resolution channel sounder is also presented, and the validity of this model is confirmed.

INTRODUCTION
In recent years, WLAN networks have been popular because of their ease of installation and mobility, and higher speed radio networks are expected. Sensor networks have also been focused as a next generation network system using radio frequencies. In these systems in which both transmitting and receiving elements are fixed, the main problem caused by channel fluctuation in indoor environments is path shadowing due to moving human bodies. The multiple-input, multiple output (MIMO) channel has been studied as a key technology for future broadband wireless networks, and path shadowing by pedestrians has a significant effect on its channel eigen values as well as the MIMO channel performance. To estimate the variation in the eigen values in a MIMO channel, the modelling of path shadowing due to human bodies will be effective. Several studies on path shadowing effects of human bodies have been reported. For example, path-shadowing probabilities for each propagation path determined by a ray tracing method were estimated to predict the received signal level at any location in a specific layout of office or factory. In it, however, the shadowing probabilities must be measured in advance along some routes human bodies move. The effects of human body shadowing on the MIMO channel performance were investigated. The influence of the human activities on a 60 GHz indoor radio channel was characterized based on measurements. These studies suggest that the influence of scattering and blocking by pedestrians is very significant on the MIMO channel performance, but the characteristics of the path shadowing probability on propagation paths has not been fully investigated yet. In this paper, a new probability model is proposed to estimate the path shadowing due to human activities. We evaluate the shadowing probabilities as well as the durations and intervals by computer simulations in order to characterize the path variation process, and propose a method for individually and independently generating the shadowing effects on each propagation path. The experimental results using a channel sounder in indoor populated environments are also presented to show the validity of the proposed model.

SIMULATION SETUP
A dielectric cylinder is commonly used as a quasi-human body model in the microwave radio channel analysis. In this paper, a dielectric cylinder model with finite height Hb = 1.7 m and diameter w = 0.4 m is used as the human body. In the computer simulations, an empty rectangular space is assumed for indoor radio propagation environments, and there is no obstruction such as furniture. Figure illustrates the simulated environment. Table I gives the parameters for a typical simulated environment. Since the aim of this computer simulation is to model the path shadowing probability due to human bodies, the constitutive parameters such as reflection coefficients are not considered, and the received level or path gain of each path is not evaluated. Several computer simulations were carried out under various conditions or parameters such as room size and ceiling height, but this paper gives several results focused on the population density.

When the rounding of a value, y, is defined as round(y), round(Sn) human bodies were generated, and randomly moved in the room. S and n indicate the area of the room and population density, respectively. Human bodies were moved every 0.1 m for each sampling time between 5,000 and 10,000. Calculations were performed to check whether each propagation path has intersections with human bodies or not out at each sampling instance. The shadowing probabilities were obtained by calculating the ratio of the number of shadowed sampling points to that of all the sampling points.

SIMULATION RESULTS
A. Path-Shadowing Probability The visibility, Pv, for the direct path between the transmitting and receiving points is given by:

(1) where x is the portion length of the propagation path that exists in the area lower than the height of human model, n and w represent the population density and width of the human bodies, corresponding to the diameter of human model, respectively. The path-shadowing probability P can be defined using (1) as:

(2) For the path reflected on ceiling/floor, the length x can b approximately

obtained by multiplying the path length, d, by the correction factor, Co, as x = d Co. Co is given as following: (3)

The following graphs show the results of computer simulations for several population densities. The lines are obtained by (4). It is found that the path shadowing probability for each path increases when the population density increases, and that (4) provides good estimations of the shadowing probabilities for different population densities.

B. Shadowing Duration and Interval This paper adopts the two-state Markov model shown in figure to characterize individual path shadowing variation due to moving human bodies. The transition probability from an unshadowed (good) state to a shadowed (bad) state is defined as Pgb, and Pbg is defined as the inverse.

The above graphs shows these state transition probabilities evaluated from the entire computer simulations conducted under many kinds of situations. The total number of evaluated paths is 22.627. This graph also shows the regression curves. The value of Pbg can be regressed by a polynomial of degree 3 for the path shadowing probability P, as follows:

(5) Based on the properties of Markov model, the Pgb drawn by a dashed line in the graph is given by

(6)

(7) This result shows that the transition probabilities can be estimated by only the path shadowing probability.

C. Modeling To generate the shadowing effects on the propagation paths, it is important to deal with the shadowing by multiple bodies. This paper assumes that the shadowing events in the computer simulations occur randomly on the time axis and follow the Poisson process. The Poisson distribution is given by:

(7) (7) gives the probability that N persons exist inside the d w rectangular space. The method to generate the shadowing events on each path is constructed by following Steps 1-4: 1. Determine propagation paths using the ray tracing method, and calculate the path length, d(i) [m], of the ith path. 2. Calculate the path shadowing probability of i-th path, p(i) , based on (4) using the path length d(i) . 3. Calculate the transition probabilities, Pbg(i) , Pgb(i) , based on (5) and (6) using the path shadowing probability p(i) . 4. Transit the states using the transition probabilities, Pbg(i) , Pgb(i) and simulate the shadowing process. When a state transits to the shadowed state, generate a random digit greater than 1 using (7), and use this number as a state bit. The number gives how many shadowed points exist on the i-th path at this state.

MEASUREMENTS OF SHADOWING CHARACTERISTICS

A. Measurement Configuration The indoor measurements were conducted using a 5.2 GHz high resolution channel sounder RUSK DoCoMo. A vertically polarized omni-directional half-wave length sleeve dipole antenna was used as the transmitting antenna, and the transmitted power was +40 dBm. The receiving antenna was a stacked polarimetric uniform circular patch array, which consists of 96 polarization sensitive single elements arranged on a cylinder in 4 rings with 24 elements each. All the elements have the vertical and horizontal polarization ports, and it is possible to measure both polarizations at almost the same time. Therefore the number of the antenna ports is 192 in total. All the receiving antenna elements have a half-power beam width (HPBW) of about 120 degrees in azimuth, and about 90 degrees in elevation. To measure the channel characteristics, a periodic multitone signal with 100 MHz bandwidth was radiated from the transmitting antenna. The maximum measured excess delay was set as 800 ns. The Time Grid mode that measures one snapshot at every 25.6 ms was used, so the sampling frequency was about 39 Hz. B. Measurements Environment The measurements were carried out in a room which size, (Lx, Ly, Lz) = (6.4, 9.6, 2.7) in meter, corresponds to the simulation environment. All the furniture was moved out from the room, and all windows were covered by aluminium plates to create an empty closed space Seven measurements were taken under three population densities. It took about 30 seconds for a measurement, thus the number of the snapshots was about 1450. C. Measurement Data Processing Measurement data was analyzed using RIMAX algorithm. The RIMAX uses two components to model the radio propagation channel. One component is based on the classical ray-optical model, which describes the radio channel by a superposition of a finite sum of concentrated (specularly-alike) propagation paths. The second component called dense multipath component (DMC) describes the contribution of distributed

diffuse scattering to the radio channel. The number of specular paths solved by RIMAX must be given as a parameter. In this study, the number was set to 10 because of the limitation of our computing resource. The information of the specular paths solved by RIMAX algorithm is following; the directions of arrival (in azimuth and elevation), time delay, and complex path weights of both vertical and horizontal polarizations. It is necessary to investigate which path data belong to the same specular path each other, because all path parameters are solved individually and independently in each snapshot, and no information for the relationship or continuity between snapshots is provided. Since the transmitting and receiving antennas were in standstill condition in this measurement, it is expected that the DoA and the time delay of each path might not change significantly. Thus the following two parameters were used for identifying specular paths over the whole snapshots; the azimuth angle of arrival and time delay. Figure below shows the estimated data for Meas. 002 plotted on azimuth-time delay plane. It can be assumed that a group of data which piles on the almost same point as shown in figure with dashed ellipse indicates to be the same paths data. The path data surrounded with dashed line ellipse shown in figure are solved almost all the snapshots, and their path parameters can be extracted easily. Next figure shows the time variation of their path weights.

In this path analysis, since the number of specular paths was set to be 10, the RIMAX algorithm might lose some desired paths which path weights becomes lower than that of the other 10 paths. This decrease of path weight might occur when the path was shadowed by human body, and thus it can be considered that data missing indicates the event with path weight becoming lower. In this study, if the lost parts are less than

seven snapshots, (< 0.18 sec), these events are assumed as just failures in solution so that the missing parts are tied up snapshots, (>0.2 sec), these events are assumed that the path weight decrease occurs due to human body shadowing. D. Measurement Result

Figure above shows the measured path shadowing probabilities. The lines in were calculated by (4) for each population density. It can be confirmed that the shadowing probability for each path increases when the population density increases, and that (4) can well estimate the measured results. Figure shows the measured transition probabilities plotted versus the shadowing probability. The transition probabilities calculated by (5) and (6) from the computer simulations have the good agreement with the measured results. From these results, it is confirmed that both of the shadowing probability and transition probability in the proposed model are valid in comparison with the measurements.

CONCLUSION
This paper presented a new probability model to estimate shadowing effects on each propagation path due to moving human bodies in indoor environments. Number of computer simulations was carried out in order to evaluate the shadowing process in indoor environments. We modelled the shadowing process as the two-state Markov model, and proposed a method to generate the shadowing effects individually and independently on propagation paths using the transition probabilities. The propagation measurements using a high resolution channel sounder were also conducted to confirm the validity of the model presented in this paper. The comparison of simulation results and measured ones concludes that the shadowing probabilities and the transition probabilities obtained from the proposed model have enough accuracy to estimate the occurrence of shadowing due to human activities in indoor environments.

REFERENCES
PAPER PRESENTED: Ippei Kashiwagi, Member, IEEE, Tokio Taga, Senior Member, IEEE, and Tetsuro Imai, Member, IEEE Time-Varying Path-Shadowing Model for Indoor Populated Environments IEEE trans. on vehicular technology, vol. 59, no. 1, January 2010 JOURNALS: S. Obayashi and J. Zander, A body-shadowing model for indoor radio communication environments, IEEE Trans. Antennas Propag., vol. 46, no. 6, pp. 920926, Jun. 1998 K. Sato and T. Manabe, Estimation of propagation-path visibility for indoor wireless LAN systems under shadowing condition by human bodies, in Proc. IEEE VTC, Ottawa, ON, Canada, May 1998, vol. 3,pp. 21092113 A. Richter and R. S. Thoma, Joint maximum likelihood estimation of specular paths and distributed diffuse scattering, in Proc. IEEE VTCSpring, Jun. 2005, vol. 1, pp. 1115 A. Richter, M. Landman, and R. S. Thoma, RIMAXA flexible algorithm for channel parameter estimation from channel sounding measurements, COST273 TD(04)045, Athens, Greece, Jan. 2628, 2004 C. C. Chong, C.-M. Tan, D. I. Laurenson, S. McLaughlin, M. A. Beach, and A. R. Nix, A novel wideband dynamic directional indoor channel model based on a Markov process, IEEE Trans. Wireless Commun., vol. 4, no. 4, pp. 1539 1552, Jul. 2005 TEXTBOOKS: Wireless Communications Principles and practice-2nd Edition, Rappaport PPT Channel Sounding - PG Course in Radio Communications -Fabio Belloni T. S.