Optimization of Access Point Placement for Automatic by ybg79195


 Optimization of Access                                            Rationale
   Point Placement for                                             Environment Description
Automatic Design of large                                          Optimization Technique
      Scale WLANs                                                  Fitness Function
                                                                   Channel Model
            Alan Mc Gibney, Martin Klepal
                                                                   Future Work
    Centre for Adaptive Wireless Systems Group,
            Cork Institute of Technology
                   Cork, Ireland

 Rationale                                                    Environment Description
 The design of wireless networks is currently still carried                                                   site-
                                                                To enable the designer to design flexible and site-specific
 out in an ad-hoc fashion, with access point installation
           ad-                                                  IEEE802.11 networks, software was developed to allow the
 based on “rules of thumb”                                      Designer to describe the environment where the proposed
                                                                WLAN will be deployed

 The objective of this project is to address the issues of     Simulation Testbed
 automatic prediction for the optimal layout of wireless
 access nodes.

 Optimization of node placements is based on

  1) Coverage Prediction              2) User Demands

 3D Multi Wall Model

Optimization Technique                      Optimization Technique

 Evolution Strategies are       used   to      To improve the performance of Optimization
 Optimize AP Placement                              pre-
                                               some pre-processing needs to be done.

                                            A) Selection of Candidate   B)     Segmentation of
                                                  Access Points              Optimization Problem

Fitness Function                            Channel Model

 FF = w1D + w2R + w3C + w4A + S                The user demand (D) is calculated by
                                               getting   the   mean    error between
 D…… User Demand Satisfaction                  estimated throughput for a target area
                                               and the predicted throughput.
 R…… Restricted Areas
 C…… Channel Assignment
                                               In order to get an accurate prediction of
 A…… Number of Access points                   throughput we needed to investigate the
 S…… Segmentation                              impact people have on the signal level.
 wi….. Weighting Factors
                            Characterisation                                                                                                                                               Characterisation
                           of Signal Variation                                                                                                                                            of Signal Variation
              Example of measured signal level
                                                                                                                                                          LOS:                                                          1 Person
     Outdoor Configuration                                                                                                                                Rice Distribution PDFRice(k)

                                                                                                                                                          Rayleigh / Lognormal Distribution
                                                                                                                                                          PDFRLN(µs, σs)

                                                                                                                                                          Overall                                                       5 People
                                                                                                                                                          PDF = A PDFRice + (1-A) PDFRLN

                                       Distribution of Signal Variation

                                                                                                                                               k          Rician k-factor (-)
                                                                                                                                               σs Standard deviation of slow fading (dB)                                14 People
                                                                                                                                               µs Mean attenuation by moving people (dB)
                                                                                                                                               A Time sharing between both states (-)
                                                                                                                                               PR Signal level with absence of moving people (dBm)

                Channel Model                                                                                                                                            Channel Parameters
                                                                     Fading                                                                                                             Outdoor Measurement                                 Indoor Measurement
                                                                     Process                                                                                                                                             NLOS Measurement
                                                            Time Share of                Switch controlled
    Rice Distribution PDFRice(k)                            Shadowing A                    by two state
                                                                                          Markov model
                                                                                                                                          Access Point

                                                         Rice        Good        Bad        Rayleigh /
    NLOS:                                               Fading
                                                                       Channel State
                                                                                            Lognormal                  µs   σs
    Rayleigh / Lognormal Distribution                   1
                                                                 +                                           10 x 20   -
    PDFRLN(µs, σs)                                                                                Fading                    Independent
                                                             1   k                               µ s, σs                      Gaussian

    Overall                                                                          Spectral
                                                                            P( f )   Shaping
    PDF = A PDFRice + (1-A) PDFRLN                                                                                                             Mean attenuation by moving                            Standard deviation of               Time sharing between
                                                                                     Rayleigh                                                  people                                                slow fading                         both states

k   Rician k-factor (-)
σs Standard deviation of slow fading (dB)
µs Mean attenuation by moving people (dB)
A Time sharing between both states (-)
PR Signal level with absence of moving people (dBm)

                                                                                                                                                         σ S (l , ρ p ) = log 7 (55lρ p + 1) + 0.5        µ S (l , ρ p ) = (3lρ p )0.7        A(l , ρ p ) = (1 − ρ p )
                                                                                                                                                                                                                                                                     0.2 l
      Site-specific Prediction of Channel
                                                                                    Impact on APs Placement

Environment Description                     Signal Level (dBm)                                     Throughput Prediction
                                                                            In Medium Populated Building
                                                                                                                     In Empty Building

                                            Additional Signal Attenuation
   Environment                              by People (dB)
   Description          PwT       freq

             Empirical model
            (Multi-Wall model)
                   L    Ι     k

               Link Level Simulator
            with Indoor Channel Model
            BER         SNR     Data Rate   Standard Deviation (dB)

                 IEEE802.11 Simulator

Future Work

Tune and evaluate optimization technique
by conducting large scale measurement in
various environments
Finalise user friendly ready-to-use tool to
provide fast and accurate WLAN design.                                            Thank you for your attention!
Aid in the development of a simulator of a
ubiquitous computing environment

To top