NYU WIRELESS The World's First Academic Research Center Combining - - PowerPoint PPT Presentation

NYU WIRELESS

The World's First Academic Research Center Combining Wireless, Computing, and Medical Applications

 T. S. Rappaport 2016

NYU WIRELESS

Investigation of Prediction Accuracy, Sensitivity, and Parameter Stability of Large-Scale Propagation Path Loss Models for 5G Wireless Communications

Shu Sun, George R. MacCartney, Jr., Theodore S. Rappaport {ss7152,gmac,tsr}@nyu.edu And co-authors Please see IEEE Trans. Vehicular Technology paper of same title

2 UMa LOS Scenario: CI: PL 𝑔, 𝑒 = 32.4 + 20𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 4.1 dB UMa NLOS Scenario: ABG: PL 𝑔, 𝑒 = 19.2 + 34𝑚𝑚𝑚10 𝑒 + 23𝑚𝑚𝑚10 𝑔 , 𝜏 = 6.5 dB CI: PL 𝑔, 𝑒 = 32.4 + 30𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 6.8 dB UMi SC LOS Scenario: CI: PL 𝑔, 𝑒 = 32.4 + 21𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 3.8 dB UMi SC NLOS Scenario: ABG: PL 𝑔, 𝑒 = 22.4 + 35𝑚𝑚𝑚10 𝑒 + 21𝑚𝑚𝑚10 𝑔 , 𝜏 = 7.8 dB CI: PL 𝑔, 𝑒 = 32.4 + 32𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 8.1 dB UMi OS LOS Scenario: CI: PL 𝑔, 𝑒 = 32.4 + 19𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 4.2 dB UMi OS NLOS Scenario: ABG: PL 𝑔, 𝑒 = 3.7 + 41𝑚𝑚𝑚10 𝑒 + 24𝑚𝑚𝑚10 𝑔 , 𝜏 = 7.0 dB CI: PL 𝑔, 𝑒 = 32.4 + 29𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 7.1 dB

Multi-Frequency Path Loss Model Performance Using Results from Industry White Paper

Note: f is in GHz and d is in meters. These forms are in 3GPP/ITU style.

• K. Haneda et al., “5G 3GPP-like channel models for outdoor

urban microcellular and macrocellular environments,” 2016 IEEE 83rd Vehicular Technology Conference (VTC Spring), May 2016. [Online]. Available: http://arxiv.org/abs/1602.07533.

• K. Haneda et al., “Indoor 5G 3GPP-like channel models for
• ffice and shopping mall environments,” 2016 IEEE

International Conference on Communications Workshops (ICCW), May 2016. [Online]. Available: http://arxiv.org/abs/1603.04079

3

Multi-Frequency Path Loss Model Performance Using Results from Industry White Paper

InH Office LOS Scenario: CI: PL 𝑔, 𝑒 = 32.4 + 17𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 3.0 dB InH Office NLOS Scenario: Single-Slope Models: ABG: PL 𝑔, 𝑒 = 17.3 + 38𝑚𝑚𝑚10 𝑒 + 25𝑚𝑚𝑚10 𝑔 , 𝜏 = 8.0 dB CIF: PL 𝑔, 𝑒 = 32.4 + 32 ∗ (1 + 0.06 ∗ (𝑔 − 24.2)/24.2) ∗ 𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 8.3 dB InH Office NLOS Scenario: Dual-Slope Models: ABG: PL 𝑔, 𝑒 = 33.0 + 17𝑚𝑚𝑚10 𝑒 + 25𝑚𝑚𝑚10 𝑔 , 1 m < 𝑒 < 6.9 m 33.0 + 17𝑚𝑚𝑚10 6.9 + 25𝑚𝑚𝑚10 𝑔 + 42𝑚𝑚𝑚10 𝑒/6.9 , 𝑒 > 6.9 m 𝜏 = 7.8 dB CIF: PL 𝑔, 𝑒 =

• 32.4 + 25 ∗ (1 + 0.12 ∗ (𝑔 − 24.1)/24.1) ∗ 𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 1 m < 𝑒 < 7.8 m

32.4 + 25 ∗ (1 + 0.12 ∗ (𝑔 − 24.1)/24.1) ∗ 𝑚𝑚𝑚10 7.8 + 20𝑚𝑚𝑚10 𝑔 + 43 ∗ (1 + 0.04 ∗ (𝑔 − 24.1)/24.1) ∗ 𝑚𝑚𝑚10 𝑒/7.8 , 𝑒 > 7.8 m 𝜏 = 7.7 dB

Note: f is in GHz and d is in meters. These forms are in 3GPP/ITU style.

• K. Haneda et al., “5G 3GPP-like channel models for outdoor urban microcellular and macrocellular environments,” 2016 IEEE 83rd

Vehicular Technology Conference (VTC Spring), May 2016. [Online]. Available: http://arxiv.org/abs/1602.07533.

• K. Haneda et al., “Indoor 5G 3GPP-like channel models for office and shopping mall environments,” 2016 IEEE International

Conference on Communications Workshops (ICCW), May 2016. [Online]. Available: http://arxiv.org/abs/1603.04079

4

Multi-Frequency Path Loss Model Performance Using Results from Industry White Paper

InH Shopping Mall LOS Scenario: CI: PL 𝑔, 𝑒 = 32.4 + 17𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 2.0 dB InH Shopping Mall NLOS Scenario: Single-Slope Models: ABG: PL 𝑔, 𝑒 = 18.1 + 32𝑚𝑚𝑚10 𝑒 + 22𝑚𝑚𝑚10 𝑔 , 𝜏 = 7.0 dB CIF: PL 𝑔, 𝑒 = 32.4 + 26 ∗ (1 + 0.01 ∗ (𝑔 − 39.5)/39.5) ∗ 𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 𝜏 = 7.4 dB InH Shopping Mall NLOS Scenario: Dual-Slope Models: ABG: PL 𝑔, 𝑒 = 22.2 + 29𝑚𝑚𝑚10 𝑒 + 22𝑚𝑚𝑚10 𝑔 , 1 m < 𝑒 < 147.0 m 22.2 + 29𝑚𝑚𝑚10 147.0 + 22𝑚𝑚𝑚10 𝑔 + 115𝑚𝑚𝑚10 𝑒/147.0 , 𝑒 > 147.0m 𝜏 = 6.4 dB CIF: PL 𝑔, 𝑒 =

• 32.4 + 24 ∗ (1 − 0.01 ∗ (𝑔 − 39.5)/39.5) ∗ 𝑚𝑚𝑚10 𝑒 + 20𝑚𝑚𝑚10 𝑔 , 1 m < 𝑒 < 110 m

32.4 + 24 ∗ (1 − 0.01 ∗ (𝑔 − 39.5)/39.5) ∗ 𝑚𝑚𝑚10 110 + 20𝑚𝑚𝑚10 𝑔 + 84 ∗ (1 + 0.39 ∗ (𝑔 − 39.5)/39.5) ∗ 𝑚𝑚𝑚10 𝑒/110 , 𝑒 > 110 m 𝜏 = 6.3 dB

Note: f is in GHz and d is in meters. These forms are in 3GPP/ITU style.

• K. Haneda et al., “5G 3GPP-like channel models for outdoor urban microcellular and macrocellular environments,” 2016 IEEE 83rd

Vehicular Technology Conference (VTC Spring), May 2016. [Online]. Available: http://arxiv.org/abs/1602.07533.

• K. Haneda et al., “Indoor 5G 3GPP-like channel models for office and shopping mall environments,” 2016 IEEE International

Conference on Communications Workshops (ICCW), May 2016. [Online]. Available: http://arxiv.org/abs/1603.04079

5

Multi-Frequency Path Loss Model Performance in the UMa Scenario

The ABG model has noticeable errors at close-in distances, i.e., it predicts much less path loss compared with free space, as well as the CI and CIF models

• S. Sun et al., ”Investigation of Prediction Accuracy, Sensitivity, and Parameter Stability of Large-Scale Propagation Path Loss Models for 5G Wireless Communications,” IEEE

Transactions on Vehicular Technology, Mar. 2016.

6

Multi-Frequency Path Loss Model Performance in the UMa Scenario

The ABG model over estimates path loss (i.e., underestimates interference) at large distances (e.g. 1 km) compared with the CI/CIF model The CI/CIF model is more conservative when analyzing interference-limited systems at large distances and more realistic when modeling signal strengths at close-in distances.

• S. Sun et al., ”Investigation of Prediction Accuracy, Sensitivity, and Parameter Stability of Large-Scale Propagation Path Loss Models for 5G Wireless Communications,” IEEE

Transactions on Vehicular Technology, Mar. 2016.

7

Multi-Frequency Path Loss Model Prediction Accuracy and Sensitivity Analysis

Shadow fading standard deviation of the ABG, CI, and CIF path loss models for prediction in distance when the prediction set is closer to the transmitter in the UMa scenario Parameters of the ABG, CI, and CIF path loss models for prediction in distance when the prediction set is closer to the transmitter in the UMa scenario

ABG: Large and unstable shadow fading standard deviation; Significant variation of model parameters CI/CIF: Small and stable shadow fading standard deviation; Little variation of model parameters

• S. Sun et al., ”Investigation of Prediction Accuracy, Sensitivity, and Parameter Stability of Large-Scale Propagation Path Loss Models for 5G Wireless Communications,” IEEE

Transactions on Vehicular Technology, Mar. 2016.

8

Multi-Frequency Path Loss Model Prediction Accuracy and Sensitivity Analysis

Shadow fading standard deviation of the ABG, CI, and CIF path loss models for prediction in distance when the prediction set is closer to the transmitter in the UMi scenario Parameters of the ABG, CI, and CIF path loss models for prediction in distance when the prediction set is closer to the transmitter in the UMi scenario

ABG: Large and unstable shadow fading standard deviation; Significant variation of model parameters CI/CIF: Small and stable shadow fading standard deviation; Little variation of model parameters

• S. Sun et al., ”Investigation of Prediction Accuracy, Sensitivity, and Parameter Stability of Large-Scale Propagation Path Loss Models for 5G Wireless Communications,” IEEE

Transactions on Vehicular Technology, Mar. 2016.

9

Multi-Frequency Path Loss Model Prediction Accuracy and Sensitivity Analysis

Shadow fading standard deviation of the ABG, CI, and CIF path loss models for prediction in distance when the prediction set is closer to the transmitter in the InH office scenario Parameters of the ABG, CI, and CIF path loss models for prediction in distance when the prediction set is closer to the transmitter in the InH office scenario

ABG: Large and unstable shadow fading standard deviation; Significant variation of model parameters CI/CIF: Small and stable shadow fading standard deviation; Little variation of model parameters

• S. Sun et al., ”Investigation of Prediction Accuracy, Sensitivity, and Parameter Stability of Large-Scale Propagation Path Loss Models for 5G Wireless Communications,” IEEE

Transactions on Vehicular Technology, Mar. 2016.

10

Conclusions

• For either single- or multi-frequency path loss models, CI/CIF models are

preferable to FI/ABG models for unknown frequencies or distances, due to their solid physical basis, fewer model parameters, frequency-dependence in the first meter, comparable shadowing standard deviation, better prediction accuracy and better parameter stability compared to the FI/ABG models

• CI model is suitable for outdoor environments, since measured path loss

exhibits little dependence on frequency in outdoor environments, beyond the first meter of free space propagation (that is captured in the CI model)

• CIF model is well suited for indoor environments, as incorporates the

frequency dependence feature of path loss observed beyond the first meter in indoor environments