Você está na página 1de 4

PROPAGATION MODLES FOR INDOOR RADIO NETWORK PLANNING INCLUDING TUNNELS

Gerd Wlfle Institut fr Hochfrequenztechnik, University of Stuttgart Pfaffenwaldring 47, D-70550 Stuttgart, Germany WWW: http://www.ihf.uni-stuttgart.de Email: woelfle@ihf.uni-stuttgart.de INTRODUCTION Radio network planning for the frequency range between 500 MHz and 5 GHz requires accurate prediction models. For large cells (macrocells) and urban scenarios (microcells) different models were already published [6]. These models are very accurate and fast [1]. But for indoor scenarios and tunnels only very simple or very time-consuming models are available. This paper presents new approaches to the modelling of the propagation in enclosed spaces. A very fast rayoptical model and a new model, especially designed for the consideration of waveguiding effects in corridors/tunnels. Three different buildings were used for the comparison of the models. A new office building with concrete walls at the University of Stuttgart (Germany), an old office building with brick walls in Vienna (Austria) and the villa of Marconi, which is a very old building (brick and wood) in Bologna (Italy) [5]. The effects especially in the long corridors of the office buildings in Stuttgart or Vienna are nearly similar to the waveguiding effects in tunnels and so these results can also be used to show the performance of the models in tunnels. Possible applications of the models are [6] 1. Mobile telephone network operators who are interested in indoor or tunnel coverage for their mobile radio networks. They want to know, where they have to place their repeaters or base stations to guarantee a sufficient coverage inside buildings or in tunnels. 2. Installation of local indoor networks like WLAN (HIPERLAN) or other computer networks. They need a very efficient planning tool to determine the best locations for their base stations (and repeaters). 3. Planning of wireless local loop (WLL) applications EMPIRICAL MODELS The simplest approach are empirical models which are based on the direct ray between transmitter and receiver. The following parameters are considered for these models [5], [6]: length of the direct ray (i.e. distance between transmitter and receiver) number (and material) of the penetrated walls attenuation offset (used for calibration) The prediction of the field strength is based on these three parameters. Therefore empirical models are very fast but they are not very accurate if the direct ray is highly attenuated [1] which might occur especially in curved tunnels. RAY OPTICAL MODELS Principle Ray-optical models are based on a determination of the rays between transmitter and receiver. Multiple reflections and diffractions are considered as well as the penetration of walls. The contribution of each ray is determined with the Fresnel equations for transmission/reflection and with the equations of the GTD/UTD for diffractions. The main disadvantage of these models is the computation time. The determination of the rays requires many hours, depending on the number of interactions considered (especially the number of diffractions increases the computation time). One of the main advantages of the ray-optical models is the consideration of multiple reflections which is very important if the waveguiding in corridors or tunnels is analysed. So the main effort in the last years was spent on the acceleration of the ray-optical models. Different approaches were presented [2] leading to acceleration factors between 2 and 10.

Preprocessing of data bases The preprocessing idea was first presented for urban scenarios [3]. But it can also be extended to indoor scenarios or arbitrary enclosed spaces [1], [6]. The basic idea of the preprocessing is the elimination of multiple identical operations. The main part of the computation time of ray optical models is spent for two problems: 1. Determination of possible rays All possible reflection and diffraction points must be determined with the image theory. Therefore the image of the transmitter related to each wall/wedge must be determined. And if multiple interactions are considered also the images of the images must be determined. During the computation of the possible reflection/diffraction points only the wall(s) considered for the reflection/diffraction is (are) analysed. Intersections between the possible new ray and all walls of the building are not considered in this step. 2. Determination of intersection with walls After the possible new ray is determined (step 1), the intersections between the ray and all walls of the building/ tunnel are computed. The ray includes as many transmissions as intersections occur. Because rays with a high number of transmissions will have a very small contribution to the total field strength, they can be neglected. To decide about the status of the ray, the transmissions (intersections) of the ray must be determined. Especially for the second step of the computation different acceleration methods were presented in the last years. [2], [4]. A list of visibility relations between the walls is generated. If a part of a wall is visible from a part of a different wall, the visibility relation between these two walls is positive and a ray might occur between these two walls without intersecting another wall. If no line of sight ray between any points on these the two walls is possible, the visibility is negative and the minimum number of transmissions is stored in the preprocessed data base. If a ray between the two walls is considered during the prediction, the minimum number of transmissions of this ray can be read from the preprocessed data and so the status of the ray can be determined without computing all intersections. The reduction of this computation effort leads to acceleration factors in the range of 2..10. But the main computation time is still spent on the determination of possible rays. And for this problem a Tile new approach was presented in [1], [6] which Segment subdivides all walls and wedges into tiles and Centre of a tile segments during the preprocessing of the data base Centre of a segment (see figure 1). The centre of each tile and segment Receiving point represents the tile/segment. The ray between the centre of each tile/segment and all further tiles/ segments of the data base is determined and stored in the preprocessed data base (including the angles of incidence/diffraction). Also the plane of the receiver pixels is subdivided into tiles and the visibility relation between these tiles and the tiles/segments of the walls are determined and stored in the preprocessed data. So a prediction is only possible for the centres of the receiver tiles, but if these tiles are Fig. 1: Subdivision of the data base and computation of visibility small, this leads to no limitation of the model. The visibility relations between the tiles and PREDICTION segments are independent of the location of the Direct ray transmitter. Only the visibility relations between the transmitter and the tiles/segments/receiver pixels 1.interaction PREPROdepend on the location of the transmitter. CESSING During the prediction only these visibility relations must be determined. All further relations can be read 2.interaction from the preprocessed data (see figure 2). This reduces the computation time for ray optical models to a few seconds/minutes on standard PCs. Further information about the algorithms, the 3.interaction computation time for the preprocessing (a few minutes up to 1 or 2 hours) and prediction (a few seconds up to 1 or 2 minutes) and the file sizes after transmitter receiving point tile / segment the preprocessing (1...30 MB) is given in [3], [6]. Fig. 2: Tree representing the interactions and rays
min max min max

DOMINANT PATH MODEL Principle An analysis of the results obtained with ray optical models shows that a high number of rays penetrates the same walls and passes the same rooms. The contributions of these rays are nearly similar. So they can be combined to a path describing the same effect as the group of individual rays. This path is called room-oriented path. The approach is presented and described in [5], [6]. The determination of these room-oriented paths is even faster than the computation of the preprocessed ray optical models (see section before), because less paths must be determined compared to the huge amount of rays in ray-optical models. Each room of the building represents a node in a logical tree. Adjacent rooms are connected via a wedge in the logical tree. Each wall between two rooms represents therefore a wedge in the tree. This logical tree including all nodes and wedges is independent of the transmitter location and can be determined in the preprocessing (an algorithm determines the rooms from the wall data [6]). The preprocessed data can then be used for all further predictions. The location and the room of the transmitter defines the root node (room) of the tree. The paths between the transmitter and the receiver can be determined very efficiently in the logical tree if the room of the receiver is determined [6]. These paths describe the sequence of rooms passed and the walls penetrated by the rays [5]. Therefore they are called roomoriented paths. The remaining task for the prediction is only the computation of the contribution of the room-oriented path to the total field strength. Therefore the number of actual rays represented by the room-oriented path should be known. Waveguiding As shown in figure 3 the waveguiding due to multiple reflections is very important because it leads to a high number of rays described by the same room-oriented path. Therefore the parameter waveguiding is introduced to describe this effect and to y describe the number of rays represented by the path. The R xT waveguiding depends on four parameters: 1. Width of the corridor/room/tunnel 2. Material of the walls (reflection loss LRi) 3. Orientation of the walls relative to the path 4. Continuity of the waveguiding Fig. 3 Waveguiding in corridors After introducing a local coordinate system with the x-axis equal to the path vector and the z-axis equal to the z-axis of the i building, the first three parameters (distance di between wall and path, reflection loss LRi, and the angle between wall and path) are analysed for each wall and combined to the contribution wi(x) of wall i at point x:

1 ( i / 45) 0 45 1 (d i ( x) / 3 m ) 0 d i 3m Ai = Di ( x ) = > 45 0 d i > 3m 0

1 (L Ri / 10 dB ) 0 L Ri 10 dB Li = L Ri > 10 dB 0 (1)

wi ( x) = Ai Di ( x) Li
The contributions of the N walls are accumulated to the guiding factor w(x), describing the guidance at a point x at the room-oriented path (see figure 3). The guiding factor w(x) is normalised with the maximum guiding factor w0:

w( x ) =

N 1 min w0 , wi ( x ) w0 i =1

(2)

The guiding factor w(x) is integrated along the path (start at the transmitter (x=xT) and stop at the receiver (x=xR)) and normalised with the path length l = xR - xT to get the total waveguiding parameter W:

1 W= x R xT

xR

x = xT

w( x ) dx

(3)

W is equal to 0 in the case of free space propagation and W is equal to 1, if an excellent waveguiding occurs. Many measurements were used to calibrate this parameter [5] and the constants are still adjusted if further measurements are available.

Further Parameters Besides the parameter waveguiding a few more parameters are necessary to describe the contribution of the roomoriented path. Among them are Length of the path Transmission loss of walls penetrated Interaction loss due to changes in the direction of the path The determination of the first and second parameter is obvious. The third parameter is computed similar to the waveguiding [6]. All parameters together describe the room-oriented path. The parameters can be determined for each part of the paths in the preprocessing of the data base, if the walls are subdivided into tiles and the parts of the paths between all tiles of a room are computed in the preprocessing [6]. So the prediction is reduced to the determination of the parts of the paths between the transmitter and the tiles of the room where the transmitter is located (same principle as for ray optical models see figure 2). If a room-oriented path has a significant contribution to the total field strength, the path is called dominant path. The computation of the field strength considers only dominant paths and not all room-oriented paths. Further information about the dominant paths is given in [5]. Prediction For each prediction point the dominant paths and their parameters are determined. Many thousands of measurement points in different types of buildings [5] were used to calibrate a regression model combining the parameters to the prediction value. Also neural networks were used to increase the accuracy and generalisation capability [6]. COMPARISON TO MEASUREMENTS Figure 4 shows the accuracy of a prediction with dominant paths for a typical indoor scenario (mean error = -0.67 dB, std-dev. = 4.64 dB). A very high accuracy is obtained with ray optical models and with the dominant paths especially in buildings with long corridors (similar to tunnels). More results and a detailed comparison of the models can be found in [5]. CONCLUSIONS Three different propagation models were presented for the planning of radio networks in enclosed spaces (tunnels/buildings). All three models have very small computation times. The empirical models are very simple and therefore not suitable for actual applications. But the rayoptical model with preprocessing and the dominant path model are not only fast, they are also very accurate. All models are implemented in a free software tool [6] and their performance and the results presented in this paper can be validated with this free demo version. REFERENCES [1] G. Wlfle, R. Hoppe, and F. M. Landstorfer, A Fast and Enhanced Ray Optical Propagation Model for Indoor and Urban Scenarios, Based on an Intelligent Preprocessing of the Database, 10th IEEE Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Osaka, Sept. 1999. [2] C. Carciofi, A. Cortina, C. Passerini, S. Salvietti, Fast Field Prediction Technique for Indoor Communication Systems, 2nd Europ. Personal and Mobile Communications Conference (EPMCC), 1997, Bonn, pp. 37-42 [3] R. Hoppe, G. Wlfle, and F. M. Landstorfer, Fast 3D Ray Tracing for the Planning of Microcells by Intelligent Preprocessing of the Database, 3rd Europ. Pers. and Mobile Communications Conf. (EPMCC), Paris, March 99 [4] G. Wlfle, F. M. Landstorfer, Intelligent ray tracing - A new approach for the field strength prediction in microcells, 47th IEEE Vehicular Technology Conference (VTC), Phoenix, May 1997 [5] G. Wlfle, P. Wertz, F. M . Landstorfer, Performance, Accuracy, and Generalization Capability of Indoor Propagation Models in Different Types of Buildings 10th IEEE Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Osaka, Japan, Sept. 1999. [6] WinProp: Software Tool for the Planning of Radio Communication Networks (Terrain, Urban, Indoor & Tunnel). Free demo-version and further information: http://www.winprop.de, Jan. 2000

Você também pode gostar