Automatic w Lan Planning

Embed Size (px)

Citation preview

  • 8/2/2019 Automatic w Lan Planning

    1/5

    Automatic Optimization Algorithms for the Planning

    of Wireless Local Area Networks

    P. Wertz, M. Sauter and F. M. LandstorferInstitut fr Hochfrequenztechnik, University of Stuttgart

    Pfaffenwaldring 47, 70569 Stuttgart, Germanywww.ihf.uni-stuttgart.de

    [email protected]

    G. Wlfle, R. HoppeAWE Communications GmbH

    Otto-Lilienthal-Str. 36, 71034 Bblingen, Germany

    [email protected]

    AbstractThe planning of wireless local area network (WLAN)infrastructures that supply large buildings or areas requires the

    consideration of many aspects (coverage, different traffic

    densities, interference, cost minimization) and therefore is a

    difficult task if done manually. In this paper a method is

    presented that allows to optimize such networks automatically.

    The approach is based on predictions of the received power toaccount for the propagation conditions that have a major impact

    on the performance of WLANs. The optimization is applied to a

    set of possible locations where access points can be installed. Out

    of this set a minimum selection of locations is made to meet the

    given requirements. These requirements consist of the

    determination of areas with different priorities and the definition

    of further parameters. The optimization not only takes into

    account the required coverage and capacity but also the

    interference situation. The arising co-channel interference is

    minimized by an appropriate assignment of the available carrier

    frequencies. The discussed approach may not find the global

    optimum in all cases, but it yields a suggestive result based on the

    locations defined by the network planner. Due to the very short

    computation time different configurations can be analyzed veryquickly.

    Keywords-Wireless Local Area Network, WLAN, Access Point,Optimization, Network Planning, Wave Propagation Modeling,

    Indoor Scenario

    I. INTRODUCTIONSince the introduction of the first wireless local area

    network (WLAN) standards, the acceptance and use ofcomponents providing high data rate connections to mobileterminals have increased very quickly. For smaller scenarioswith only a few access points (AP) to be installed, no complexnetwork planning is needed. However, network solutionssupplying larger areas like public hot-spots, university campus,office buildings etc. need much more sophisticated planningmethods [1, 2].

    At first glance the task of a network planner is to build upan infrastructure which offers sufficient coverage to fulfill thegiven requirements concerning the capacity demands. Due tothe low number of available non-overlapping frequencychannels (especially with standards using the 2.4 GHz band),the problem of co-channel interference has a major impact onthe performance of the network and should therefore also beconsidered. This results in the need to also include the carrier

    assignment in the planning process depending on the standardused for the network installation.

    As a basis for the optimization, the usage of accuratepropagation models [3, 4] is crucial (see section II). Suchmodels are available in an application [5] that also hosts the

    additional tool discussed here. This tool supports the networkplanner very effectively as it allows to find an optimizedselection of AP locations automatically in a very short time.This resulting selection will be a tradeoff between best received

    power and minimum interference which may be weighted bythe user.

    II. DESCRIPTION OF THE APPROACHIn order to find a globally optimized solution all possible

    combinations of selected APs would have to be assessed inorder to choose the one which performs best according to therequirements.

    With the number of potential locations, this combinatorialproblem leads to an exponential increasing variety of solutions.The required computation time for such an approach wouldtherefore not be acceptable, especially for the planning of largeinstallations.

    Therefore a different approach was made to find adequatesolutions with a much smaller computational demand.

    The applied algorithms are based on a set of inputparameters which allow the network planner to define thedesired requirements. The most important information is the

    position of the possible installation sites. For each of thesepotential positions, a wave propagation prediction (see figure1) is computed. As the optimization results heavily depend on

    the accuracy of the wave propagation prediction, a veryaccurate ray tracing model is used [3, 5].

    Fig. 1: Prediction of the received power using a ray tracing model

  • 8/2/2019 Automatic w Lan Planning

    2/5

    Since it is very difficult for network planners to exactlyestimate the expected traffic amount for the individuallocations in the scenario, the requirements for capacity andavailability are represented as priorities in a priority-map (seefigure 5 in section III).

    The user additionally has to define the followingparameters both for the lowest and for the highest priority:

    minimum received power carrier to interference ratio (CIR) transmitter densities (described by a minimum needed

    mean difference between the received power of twotransmitters)

    Furthermore, two different carrier assignment algorithmscan be selected and the weighting of the received power andinterference assessment in the objective function can bedefined.

    In the first step the predicted coverage of every potentiallyinstalled transmitter is assessed regarding the priority and the

    received power of each pixel in the considered area. Theassessment is realized by a piecewise defined function whichresults from a previous linguistic description (see figure 4). Themean assessment over the whole scenario is a measure for theimportance of each possible site.

    In the second step the density of the available transmitterlocations is reduced depending on the priority settings of theaffected areas. This is realized by finding groups of sites whose

    provided coverage show large overlapping, which means thatthe mean absolute difference between their provided coverageis below a certain level (see figure 2). Out of such a group theAP with the greatest importance is chosen f or furtherconsiderations as well as those APs which do not belong to any

    group. This first pre-selection algorithm provides a minimizednumber of used APs as well as better conditions to reduceinterference problems.

    Fig. 2: Dependence of power difference on priority

    In the next more costly step the remaining APs aresequentially added (following their importance) to a solutionset. Hence APs which supply important areas belong to thesolution set for sure. To each of these sets an iterative carrierassignment algorithm is applied following the steps:

    1. Start at the most relevant AP.2. Continue at the next AP from the current solution set

    that has the lowest path loss to the current AP (seefig. 3).

    carrier assigned

    carrier assigned

    assign carrier

    Fig. 3: Finding the next AP for carrier assignment

    Every pixelis assessed!

    highrelevance

    lowrelevance

    High

    lowrelevance

    Very lowrelevance

    low

    highlowprio

    power

    highrelevance

    lowrelevance

    High

    lowrelevance

    Very lowrelevance

    low

    highlowprio

    power

    max max

    0 0

    pixels in building

    ( , )

    (Site)

    x y

    x y

    CoverageAssessment x y

    CoverageAssessmentn

    = =

    =

    Parameters:

    - max. needed received power- min. needed received power

    Fig. 4: Coverage assessment

  • 8/2/2019 Automatic w Lan Planning

    3/5

    3. Assign carrier according to one of two algorithmsexplained below.

    The Received Power Algorithm distributes theavailable carrier frequencies based on the mutualreceived powers between the APs. For this, thereceived power of the APs to which a carrier wasalready assigned is evaluated at the location of theindividual AP. The carrier which is received with the

    lowest power is then chosen.

    The Mutual Interference Algorithm takes intoaccount the whole area of the scenario by assessing the

    possibly resulting interference between the APs ifworking on the same carrier frequency. Thisassessment works similar to the one for the coveragedescribed above.

    The resulting entire coverage and the co-channelinterference for each combination of APs is assessed regardingthe given priority requirements. For example areas with a high

    priority should be served by more APs with a higher receivedpower and not suffer from a bad co-channel interference.

    Therefore the already described assessment functions are used.The final objective function is the sum of both assessmentsweighted in accordance with the user settings:

    cir cir cov covOF w A w A= + (1)

    cirw and covw are the weighting factors for the assessment

    functions cirA and covA for the CIR and the coverage

    respectively.

    Finally the set of APs with the best objective function is thedesired solution and is therefore output to the user.

    III. RESULTSTo present the achievable results the scenario shown in

    figure 5 is investigated. It consists of an office building with anarea of approx. 10,000 square meters. 57 possible AP locationswere defined. Out of a range of 6 priority values (0-5) only 0(white, no WLAN access expected), 1 (green), 3 (blue) and 5(yellow, high importance) are used to mark areas with differentrequirements. The most important of the chosen parametersare:

    cir cov

    w / w 0.3 / 0.7=

    mean received power difference at priority 5 : 5 dB mean received power difference at priority 1 : 15 dBAs can be seen in figure 6, the finally selected APs provide

    a coverage which obviously depends on the entered priorities.Areas marked as unimportant (priority 0) are much more likelyto be supplied with less or insufficient received power.

    Prio 0

    Prio 1

    Prio 3

    Prio 5

    Fig. 5: Priority map including the possible access point locations

    Fig. 6: Coverage with the automatically selected APs

  • 8/2/2019 Automatic w Lan Planning

    4/5

    Figure 7 shows the resulting carrier assignment. In this casethe Mutual Interference Algorithm was applied.

    Figure 8 shows the co-channel interference situation in theentire building. The rooms with the highest priority are notaffected by interference problems. In lower priority areas someinterference can be tolerated. The worst values (blue color)occur at places with the lowest priority. Obviously the foundsolution does not prevent arising co-channel interference

    problems in any case. This results especially from the use ofonly 3 carrier frequencies. But it is evident that the interferencesituation is successfully optimized in areas with high priority(see figure 5).

    To analyze the impact of the priority map a differentexample is used where extreme priority values were defined(see figure 9). A high received power and thus good coverageis found in the areas marked with high priority

    Fig. 7: Applied carrier assignment

    Fig. 8: Resulting co-channel interference

    Prio 0

    Prio 5

    Fig. 9: Impact of the priority map

  • 8/2/2019 Automatic w Lan Planning

    5/5

    The computation time for the presented results isapproximately 5 seconds for each scenario using a standard PCwith a 1 GHz processor. The additional computation time forthe propagation analysis is in the range of 15 seconds perpossible AP site, but has to be done only once per scenario.

    IV. OUTLOOKSo far, the approach can only be applied to systems withnon-overlapping carriers. It could be extended in the way that

    the interference of overlapping carriers is adequately assessedallowing to use more of the available carriers (especially instandards working in the 2.4 GHz Band). It is expected that thisleads to a better optimization result due to the fact that there isa higher degree of freedom.

    Furthermore, the interference assessment could be extendedto consider also the uplink. This could be realized with astatistical user distribution model that again takes into accountthe priority map.

    V. CONCLUSIONSIn this paper a new approach is described to optimize

    WLANs in regard of coverage, interference and requiredavailability. The proposed approach may not find the globaloptimum that would be theoretically found by testing allpossible solutions (which is impossible in an acceptableamount of time), but it will lead to an adequate solution. Forthis, it may be necessary for the network planner to vary

    planning parameters and check the corresponding results.However, due to the short computation time, this is not a majordrawback.

    As the host application [5] also supports the propagationanalysis with a transition between urban scenarios andscenarios in enclosed spaces [6], the approach can also be usedfor the planning of networks in mixed urban/indoorenvironments (like a campus).

    REFERENCES

    [1] M. Hein, B. Maciejewski, Wireless LAN, Funknetze in der Praxis,Franzis Verlag, 2002

    [2] M. Unbehaun, M. Kamenetsky, The evolution of wireless LANs andPANs - On the deployment of picocellular wireless infrastructure, IEEEWireless Communications, Volume: 10, Issue: 6, pp. 70 80, Dec. 2003

    [3] R. Hoppe, G. Wlfle, P. Wertz, F. M. Landstorfer, Advanced Ray-Optical Wave Propagation Modeling for Indoor Environments IncludingWideband Properties, European Wireless 2002, Firence (Italy), Feb.2002

    [4] C. Carciofi, A. Cortina, C. Passerini, and S. Salvietti, Fast FieldPrediction Techniques for Indoor Communication Systems, 2nd

    European Personal and Mobile Communications Conference (EPMCC),Bonn, pp. 37 - 42, Nov. 1997

    [5] AWE Communications, Germany, Software Tool WinProp for thePlanning of Mobile Communication Networks (incl. demo-version),www.awe-communications.com, January 2004

    [6] P. Wertz, D. Zimmermann, R. Hoppe, G. Wlfle, F. M. Landstorfer,Hybrid Ray Optical Models for the Penetration of Radio Waves intoEnclosed Spaces, VTC Fall 2003 , Orlando (Florida, USA), Oct. 2003