| The Creation & Use of Imagery for Cellular Network Planning
By Alex Zaloumis Abstract
Aerial photography, satellite or raster/vector combinations of imagery are considered traditional (standard) datasets, while a more user-defined dynamic type of image is becoming more common. Traditional imagery is still used as a background template for the distribution of both internal and external information directly to customers.
The purpose of dynamic imagery is to reflect a current situation in a cost-effective and responsive manner. This imagery often combines both external- and internal-captured information.
It is a continuous exercise to consistently update dynamic imagery with current information. This ensures the long-term validity and usefulness of the imagery. The combination of traditional and dynamic imagery results in a dataset that aids decision-making. This dataset is also understandable because of its spatial context. The user-friendliness of the data means that additions and updates are both encouraged and easily incorporated.  1:20,000 aerial photography was ortho-rectified and geo-referenced so that a custom 10-meter DEM could be generated. Due to the cost of such an exercise, this process was limited to urban areas (Figure 1).
Introduction
In 1994, Mobile Telephone Networks (MTN) was granted a license as South Africa's second national GSM cellular provider. The company planned and built the entire network. Spatial data was used to determine the deployment and layout of the physical network. Data consisting of DEM, clutter (a type of land-use classification) and vectors were used to position sites based upon propagated signal coverage as determined by the slope of the land, engineering modeling parameters, and similar factors.
At MTN the emphasis moved from providing coverage to ensuring network capacity. Standard sets of imagery, even the most current, usually provide only a snapshot view and do not reflect overall network performance. In order to achieve newly established goals, and to model an existing network, dynamic imagery is required. This imagery is derived from continuously captured network information and statistics, and then combined with traditional datasets for purposes of analysis and evaluation. The Overall Process
Phase One is concerned with the deployment of the sites in the shortest possible timeframe, and in a manner that will cover the greatest amount of area with the best possible signal quality.
Phase Two concentrates on monitoring the finished network. This is done for two reasons, first to identify holes in the rollout, and second to assess the quality of service, both from a technical and a customer-expectation point of view. Performance data collected from the network is mostly tabular, statistical information, providing a two-dimensional view of a three-dimensional operation.
Phase Three, often the final one, is the expansion and improvement of the network, a partly cyclical exercise in conjunction with Phase Two. Inputs at this stage include socio-economic-demographic data and the personal knowledge of the users. It is this final input that contributes indeterminable value to the analysis, and is the hardest to identify and capture. Phase One: Utilizing Traditional Imagery Datasets
To determine the positioning of the physical network (sites, switches, etc.) four primary spatial datasets are used. These are vector, digital elevation model (DEM), clutter, and raster backdrops. Vector data is comprised of line-and-point objects such as roads, rivers, police stations, etc. This imagery is used primarily for location and reference purposes. For this project the data was acquired from both commercial data mapping companies and the South African national mapping agency. The vectors had been captured using a variety of techniques such as field surveys, heads-up digitizing, and aerial photography.
Both DEM and ground terrain affect signal coverage and, therefore, affect the design and the positioning of cell sites or towers. Since a consistent terrain model remains one of the most difficult of all datasets to obtain in South Africa, color 1:20,000 aerial photography was ortho-rectified and geo-referenced so that a custom 10-meter DEM could be generated. Due to the cost of such an exercise, this process was limited to urban areas. (Figure 1)
At present, South Africa has a 20-meter DEM that was generated from the vectorization of the CDSM 1:50,000 map series. These vectors were then interpolated to create a raster product. The process relies on the accuracy of the maps, and no slide-rule check exists to verify the accuracy of the resultant DEM.
Another method for deriving DEM was radar technology that uses differential GPS. Sub-meter accuracy is achievable, although only after a certain amount of post-processing. One of the benefits to this method is that raw-data capture is not dependent upon weather, thus allowing greater flexibility. So far this method has not been tried.
Clutter is a type of land-use classification. It varies from the traditional set of land-use data in that its main criteria for identification are feature height, feature density, and feature texture, such as broad-leaf versus small-leaf forestry. These three criteria affect both the strength of the signal and the distance it can travel.
The clutter was generated from the same aerial photography that was utilized for the DEM generation in the urban areas. Vector polygons were captured via heads-up digitizing, with a class identifier assigned to each object. Then the vector layer was cleaned and checked for overshoot errors, slivers, and so forth. No re-projecting was required as the vector layer was referenced to aerial photography, and then rastorized. Raster-clutter resolution was then doubled to 20 meters due to the practical considerations of disk space and the speed of loading applications.
The final set of standard imagery used was the background (scanned backdrop) image. These were mostly original 1:20,000 aerial photographs with a combination of medium-scale satellite images to fill the gaps in largely rural areas. The engineers used this imagery primarily for location purposes and feature identification, for example, to plot the surface for mast placement.
Still photographs and videography of the mast were also used as a type of background imagery. Still photography included panoramic views of the site. Videography was used to identify surrounding textures and the placement of the site, for instance, what the exact height of the mast should be, and how difficult it was to position. Phase Two: Creating Dynamic Imagery; Combining with Traditional Imagery
Once the network was built, it was then possible to capture statistical information on the network's performance. Customer information included the number of subscribers on the network, where they roamed, the duration of their calls, what type of handsets they used, and so on. One result of capturing this data was to identify areas where services were not being delivered, yet were actually required.
Customer information was represented through the site's signal coverage area, namely a polygon. (Figure 2) This was generated in the planning tool, which uses the traditional data to generate a signal-coverage polygon. This polygon was then exported into a mapping tool and, when combined with the dynamic statistical imagery, accurately shows subscribers' activities within a signal-coverage area.
Information on sites and switches included information such as congestion, frequency behavior, routing performance, which sites were carrying the most traffic, and which sites performed the best technically.
Site statistics are geographical point data. Thematic imagery created from this data can be used to indicate up-to-the-minute network alarms (site failure, etc.) and other potential problems that might include call congestion during holiday time at vacation sites. Trending of specific areas over a period of time is also possible to prevent the same problems from cropping up again.
The difficulty with this part of the process is that statistical data is captured in systems that output only raw tabular information. One constraint is the time and effort required for the extraction and conversion of the data into conventional mapping and GIS tools. Once the statistical data has been imported into the mapping tool and combined with the standard datasets, only then can the benefits of this dynamic imagery be seen. Phase Three: Combining and Overlaying All Possible Imagery Sets
At this stage the traditional and dynamic imagery were overlaid, and the consistencies between similar natural habitats and similar network/subscriber patterns became visible. Where such patterns did not emerge, further analysis was done to identify why the patterns were not as expected. A third dimension, demographic-socio-economic (DSE), was then added.
Demographics, poverty gaps, age groups, and employment ratios are but a few criteria that sometimes identify and explain behavior patterns that occur on the network. This data is used only at this point, mainly because of a past history of unreliability and age. Instead of taking these DSE data at face value, true figures - network statistics and standard data sets - are added to the DSE information, and trends can then be identified.
Certain assumptions about DSE information can be made, based upon investigations via other sources. Where these match results obtained from the network itself, the DSE information is considered fairly reliable. One future project is to identify profitable areas, and then investigate network growth and expansion into other areas that have the same or similar profile. User Interactive Imagery
In order for imagery to assist with proactive decision-making, it must be a combination of standard imagery and dynamic information and, when identified as necessary, DSE imagery as well. Another reason why DSE imagery is left for last is because of the cost of acquiring such data, especially when it is more recently dated.
A possible input at this stage is the spatial knowledge of the user, that is, the knowledge of the field. The problem is in trying to capture this "wet" information into the mainstream data and imagery. MTN's solution was designed to address two issues. The first was how to extract data from the users in a simple and time-effective manner. Second, the data was then deployed as information to other users, not simply viewed and then stored away for future reference.
It was then decided to disperse this imagery via an intranet. Access to the data at varying levels of interactivity and contribution was granted depending upon the status of the user. Individual decision-making and analysis was done using queries and filtering techniques, while allowing for inputting of personal knowledge, which was then committed to the database once its value and accuracy had been determined by management. Conclusion
This article has illustrated how standard imagery sets are now complimented and altered by the use of internal dynamic imagery and external DSE information. Imagery is comprised of both the universal picture and personalized user-pertinent information. This combined image becomes a tool that can be used to assist with decision-making and overall strategic planning. About the Author:
Ms. P.A. Zaloumis has been practicing GIS professionally for the past four years, focused primarily on the cellular telephone industry. Among her responsibilities are the sourcing and preparation of datasets for radio engineers, management and maintenance of various datasets, and GIS analysis for cellular planners, the sales and marketing department, and implementation engineers. She may be reached at Mobile Telephone Networks (MTN), Private Bag 9955, Sandton (2146) Rep. Of South Africa. Tel: (27) 11-301-6000. E-mail: [email protected] Editors Note: This material was originally published in the GIS 2001 conference proceedings published by GeoTec Media, www.gisconference.com Back |
