Modeling Environmental Effects on Directionality in Wireless - - PowerPoint PPT Presentation

modeling environmental effects on directionality in
SMART_READER_LITE
LIVE PREVIEW

Modeling Environmental Effects on Directionality in Wireless - - PowerPoint PPT Presentation

Modeling Environmental Effects on Directionality in Wireless Networks Eric Anderson, Caleb Phillips, Douglas Sicker, and Dirk Grunwald eric.anderson@colorado.edu University of Colorado Department of Computer Science 26 June 2009 Eric


slide-1
SLIDE 1
  • Modeling Environmental Effects on Directionality in

Wireless Networks

Eric Anderson, Caleb Phillips, Douglas Sicker, and Dirk Grunwald eric.anderson@colorado.edu

University of Colorado Department of Computer Science

26 June 2009

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 1 / 22

slide-2
SLIDE 2
  • Outline

1

Radio Propagation Environments and Directional Antennas Pretty Pictures Measuring Effective Directionality Accuracy of Current Models

2

Estimating Radio Propagation Ray Tracing Propagation Models Directivity Models

3

Directivity and Propagation are not Orthogonal What’s Missing?

4

Modeling Fitting Specific Environments Fitting Types of Environments Relating Signals to Environment Parameters

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 2 / 22

slide-3
SLIDE 3
  • RF Propagation Environments

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 3 / 22

slide-4
SLIDE 4
  • RF Propagation Environments

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 3 / 22

slide-5
SLIDE 5
  • Measurement Processes [1/2]

Baseline Measurements

Calibration and test quality equipment (Agilent E4438C, 89600S VSG and VSA) used for: Reference pattern Calibrating laptop measurements.

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 4 / 22

slide-6
SLIDE 6
  • Measurement Processes [2/2]

Laptop Measurements

Dell laptops Atheros AR5213 radios Used for non-reference measurements.

−80 −60 −40 −20 −80 −60 −40 −20

Linear fit for RSS Error

Tx dBm Rx dBm

Linear fit, slope ≈ 0.95 Adjusted R2 = 0.989

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 5 / 22

slide-7
SLIDE 7
  • How Bad is it?

25 50 75 105 140 175 210 245 280 315 350 −50 −40 −30 −20 −10 10 Patch−Panel Antenna Angle, Degrees Counterclockwise dB Relative to Peak Mean

Patch−Outdoor−B Patch−Outdoor−A Patch−Indoor−A Reference

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 6 / 22

slide-8
SLIDE 8
  • How Bad is it?

25 50 75 105 140 175 210 245 280 315 350 −50 −40 −30 −20 −10 10 Patch−Panel Antenna Angle, Degrees Counterclockwise dB Relative to Peak Mean

Patch−Outdoor−B Patch−Outdoor−A Patch−Indoor−A Reference

≥ 15 dB

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 6 / 22

slide-9
SLIDE 9
  • How Bad Is It?

21 50 76 106 140 171 205 236 270 301 335 −50 −40 −30 −20 −10 10 24dBi Parabolic Dish, Indoors Angle, Degrees Counterclockwise dB Relative to Peak Mean

Parabolic−Indoor−C Parabolic−Indoor−B Parabolic−Indoor−A Reference

≥ 35 dB

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 7 / 22

slide-10
SLIDE 10
  • Outline

1

Radio Propagation Environments and Directional Antennas Pretty Pictures Measuring Effective Directionality Accuracy of Current Models

2

Estimating Radio Propagation Ray Tracing Propagation Models Directivity Models

3

Directivity and Propagation are not Orthogonal What’s Missing?

4

Modeling Fitting Specific Environments Fitting Types of Environments Relating Signals to Environment Parameters

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 8 / 22

slide-11
SLIDE 11
  • Radio Ray Tracing

Prx = Ptx

  • fa(θ1)fb(θ2)d−2

1

  • Eric Anderson (CU Boulder)

Modeling Environmental Effects . . . WiNMee ’09 9 / 22

slide-12
SLIDE 12
  • Radio Ray Tracing

Prx = Ptx

  • fa(θ1)fb(θ2)d−2

1

+ fa(θ3)fb(θ4)d−2

2 A2

  • “two-ray”

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 9 / 22

slide-13
SLIDE 13
  • Radio Ray Tracing

Prx = Ptx

  • fa(θ1)fb(θ2)d−2

1

+ fa(θ3)fb(θ4)d−2

2 A2 + fa(θ5)fb(θ6)d−2 3 A3

  • Eric Anderson (CU Boulder)

Modeling Environmental Effects . . . WiNMee ’09 9 / 22

slide-14
SLIDE 14
  • Radio Ray Tracing

Prx = Ptx

  • fa(θ1)fb(θ2)d−2

1

+ fa(θ3)fb(θ4)d−2

2 A2 + fa(θ5)fb(θ6)d−2 3 A3 · · ·

  • Eric Anderson (CU Boulder)

Modeling Environmental Effects . . . WiNMee ’09 9 / 22

slide-15
SLIDE 15
  • Path Loss

Path loss: Macro-scale function of position & terrain. e.g. Free space, two-ray, HATA/COST231, ITU238, . . .

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 10 / 22

slide-16
SLIDE 16
  • Fading

Frequency Time [Credit: Public domain image from Wikimedia commons]

Fading: Micro-scale function of many positions and velocities. Treated as function of time. e.g. Rayleigh, Rician, Weibull, . . .

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 11 / 22

slide-17
SLIDE 17
  • Directivity – Current Models

Fading & path loss Node a gain Node b gain Prx = Ptx ∗ X ∗ fa(θ1) ∗ fb(θ2)

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 12 / 22

slide-18
SLIDE 18
  • Directivity – Current Models

Fading & path loss Node a gain Node b gain Prx = Ptx ∗ X ∗ fa(θ1) ∗ fb(θ2)

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 12 / 22

slide-19
SLIDE 19
  • Directivity – Current Models

Fading & path loss Node a gain Node b gain Prx = Ptx ∗ X ∗ fa(θ1) ∗ fb(θ2)

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 12 / 22

slide-20
SLIDE 20
  • Directivity – Current Models

Fading & path loss Node a gain Node b gain Prx = Ptx ∗ X ∗ fa(θ1) ∗ fb(θ2)

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 12 / 22

slide-21
SLIDE 21
  • Outline

1

Radio Propagation Environments and Directional Antennas Pretty Pictures Measuring Effective Directionality Accuracy of Current Models

2

Estimating Radio Propagation Ray Tracing Propagation Models Directivity Models

3

Directivity and Propagation are not Orthogonal What’s Missing?

4

Modeling Fitting Specific Environments Fitting Types of Environments Relating Signals to Environment Parameters

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 13 / 22

slide-22
SLIDE 22
  • Directivity – What’s Missing?

+

Frequency Time

+ = ? Some f(position) ∗ Some f(time)

  • Prx = Ptx ∗ X ∗ fa(θ1) ∗ fb(θ2)

vs. Prx = Ptx∗

     fa(θ1) ∗ fb(θ2) ∗ d−2

1

+ fa(θ3) ∗ fb(θ4) ∗ d−2

2

∗ A2+ fa(θ5) ∗ fb(θ6) ∗ d−2

3

∗ A3+ · · ·      

  • Eric Anderson (CU Boulder)

Modeling Environmental Effects . . . WiNMee ’09 14 / 22

slide-23
SLIDE 23
  • Directivity – What’s Missing?

+

Frequency Time

+ = ? Some f(position) ∗ Some f(time)

  • Prx = Ptx ∗ X ∗ fa(θ1) ∗ fb(θ2)

vs. Prx = Ptx∗

     fa(θ1) ∗ fb(θ2) ∗ d−2

1

+ fa(θ3) ∗ fb(θ4) ∗ d−2

2

∗ A2+ fa(θ5) ∗ fb(θ6) ∗ d−2

3

∗ A3+ · · ·      

  • Eric Anderson (CU Boulder)

Modeling Environmental Effects . . . WiNMee ’09 14 / 22

slide-24
SLIDE 24
  • Directivity – What’s Missing?

+

Frequency Time

+ = ? Some f(position) ∗ Some f(time)

  • Prx = Ptx ∗ X ∗ fa(θ1) ∗ fb(θ2)

vs. Prx = Ptx∗

     fa(θ1) ∗ fb(θ2) ∗ d−2

1

+ fa(θ3) ∗ fb(θ4) ∗ d−2

2

∗ A2+ fa(θ5) ∗ fb(θ6) ∗ d−2

3

∗ A3+ · · ·      

  • Antenna gain in “off” directions is

ignored!

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 14 / 22

slide-25
SLIDE 25
  • An Obvious Problem
  • Eric Anderson (CU Boulder)

Modeling Environmental Effects . . . WiNMee ’09 15 / 22

slide-26
SLIDE 26
  • An Obvious Problem
  • fa(θ1)

fb(θ2)

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 15 / 22

slide-27
SLIDE 27
  • An Obvious Problem
  • fa

( θ3 ) f

b

( θ

4

)

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 15 / 22

slide-28
SLIDE 28
  • An Obvious Problem
  • Eric Anderson (CU Boulder)

Modeling Environmental Effects . . . WiNMee ’09 15 / 22

slide-29
SLIDE 29
  • Outline

1

Radio Propagation Environments and Directional Antennas Pretty Pictures Measuring Effective Directionality Accuracy of Current Models

2

Estimating Radio Propagation Ray Tracing Propagation Models Directivity Models

3

Directivity and Propagation are not Orthogonal What’s Missing?

4

Modeling Fitting Specific Environments Fitting Types of Environments Relating Signals to Environment Parameters

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 16 / 22

slide-30
SLIDE 30
  • Environment Fitting

25 50 75 105 140 175 210 245 280 315 350 −50 −40 −30 −20 −10 10 Patch−Panel Antenna Angle, Degrees Counterclockwise dB Relative to Peak Mean

Patch−Outdoor−B Patch−Outdoor−A Patch−Indoor−A Reference

25 55 85 120 160 200 240 280 320 −30 −10 10 30

Orthogonal Model Error

Azimuth Angle Normalized Gain, dB

“Environment” = “Error”

Error is correlated with angle Fit existing model error

Bin by angle Find regression fit

Characterizes specific link, antenna, environment.

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 17 / 22

slide-31
SLIDE 31
  • Environment Fitting

25 50 75 105 140 175 210 245 280 315 350 −50 −40 −30 −20 −10 10 Patch−Panel Antenna Angle, Degrees Counterclockwise dB Relative to Peak Mean

Patch−Outdoor−B Patch−Outdoor−A Patch−Indoor−A Reference

25 55 85 120 160 200 240 280 320 −30 −10 10 30

Orthogonal Model Error

Azimuth Angle Normalized Gain, dB

“Environment” = “Error”

Error is correlated with angle Fit existing model error

Bin by angle Find regression fit

Characterizes specific link, antenna, environment.

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 17 / 22

slide-32
SLIDE 32
  • Environment Fitting

25 50 75 105 140 175 210 245 280 315 350 −50 −40 −30 −20 −10 10 Patch−Panel Antenna Angle, Degrees Counterclockwise dB Relative to Peak Mean

Patch−Outdoor−B Patch−Outdoor−A Patch−Indoor−A Reference

25 55 85 120 160 200 240 280 320 −30 −10 10 30

Orthogonal Model Error

Azimuth Angle Normalized Gain, dB

“Environment” = “Error”

Error is correlated with angle Fit existing model error

Bin by angle Find regression fit

Characterizes specific link, antenna, environment.

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 17 / 22

slide-33
SLIDE 33
  • Environment Fitting

25 50 75 105 140 175 210 245 280 315 350 −50 −40 −30 −20 −10 10 Patch−Panel Antenna Angle, Degrees Counterclockwise dB Relative to Peak Mean

Patch−Outdoor−B Patch−Outdoor−A Patch−Indoor−A Reference

25 55 85 120 160 200 240 280 320 −30 −10 10 30

Orthogonal Model Error

Azimuth Angle Normalized Gain, dB

“Environment” = “Error”

Error is correlated with angle Fit existing model error

Bin by angle Find regression fit

Characterizes specific link, antenna, environment.

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 17 / 22

slide-34
SLIDE 34
  • Error Comparison

25 55 85 120 160 200 240 280 320 −30 −10 10 30

Orthogonal Model Error

Azimuth Angle Normalized Gain, dB 25 55 85 118 156 195 233 271 310 348 −30 −10 10 30

Residual Error After Offset Fitting

Azimuth Angle Error in dB

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 18 / 22

slide-35
SLIDE 35
  • Types of Environments

Environment = set of offsets

25 50 75 105 140 175 210 245 280 315 350 −50 −40 −30 −20 −10 10 Patch−Panel Antenna Angle, Degrees Counterclockwise dB Relative to Peak Mean

Patch−Outdoor−B Patch−Outdoor−A Patch−Indoor−A Reference

25 55 85 120 160 200 240 280 320 −30 −10 10 30

Orthogonal Model Error

Azimuth Angle Normalized Gain, dB

e.g. {-1,-2,-5,-2,· · · } Group of environments = distribution of offsets

Offset ANOVA

Variation is predicted by: “Type” of environment Antenna gain f (θk) Observation point (negligible)

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 19 / 22

slide-36
SLIDE 36
  • Types of Environments

Environment = set of offsets

25 50 75 105 140 175 210 245 280 315 350 −50 −40 −30 −20 −10 10 Patch−Panel Antenna Angle, Degrees Counterclockwise dB Relative to Peak Mean

Patch−Outdoor−B Patch−Outdoor−A Patch−Indoor−A Reference

25 55 85 120 160 200 240 280 320 −30 −10 10 30

Orthogonal Model Error

Azimuth Angle Normalized Gain, dB

e.g. {-1,-2,-5,-2,· · · } Group of environments = distribution of offsets

Offset ANOVA

Variation is predicted by: “Type” of environment Antenna gain f (θk) Observation point (negligible)

Therefore

(Type, antenna gain) ⇒

  • ffset distribution

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 19 / 22

slide-37
SLIDE 37
  • Environment Parameters

Antenna a, bin k

Arc of azimuth centered at θk. Fitted offset is Offk E[Offk] = fa(θk) ∗ gain coefficient Offk ∼ Nor(E[Offk], σ(offset)) Signal ∼ Nor(Offk, σ(signal))

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 20 / 22

slide-38
SLIDE 38
  • Environment Parameters

Antenna a, bin k

Arc of azimuth centered at θk. Fitted offset is Offk E[Offk] = fa(θk) ∗ gain coefficient Offk ∼ Nor(E[Offk], σ(offset)) Signal ∼ Nor(Offk, σ(signal)) Environment Kgain Soff Sss Open Outdoor 0.01 - 0.04 1.326 - 2.675 2.68 - 3.75 Urban Outdoor 0.15 - 0.19 2.244 - 3.023 2.46 - 2.75 LOS Indoor 0.25 - 0.38 2.837 - 5.242 2.9 - 5.28 NLOS Indoor 0.67 - 0.70 3.17 - 3.566 3.67 - 6.69

Table: Summary of Regression Results: Gain-offset regression coefficient (Kgain),

  • ffset residual std. error (Soff ), and signal strength residual std. error (Sss).

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 20 / 22

slide-39
SLIDE 39

Thank you

Contact: eric.anderson@colorado.edu Measurements: http://www.crawdad.org/cu/antenna Simulation software (Qualnet 4.5.1 patch): http://systems.cs.colorado.edu/

slide-40
SLIDE 40
  • Bin Sizes

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 2 3 4 5 6 7 8 9 Parabolic−Outdoor−A Parabolic−Indoor−A Patch−Outdoor−A Parabolic−Outdoor−B Patch−Outdoor−B Patch−Indoor−A Parabolic−Indoor−B Parabolic−Indoor−C Patch−Indoor−B Patch−Indoor−C Array−Outdoor−A

Residual Error Relative to Bin Count (Individual Data Sets Overlaid on Box Plot) Number of Fitting Bins (Azimuth) Residual Std. Error, dB

Eric Anderson (CU Boulder) Modeling Environmental Effects . . . WiNMee ’09 22 / 22