[Back]

L. Eller, P. Svoboda, M. Rupp:
"Propagation-aware Gaussian Process Regression for Signal-Strength Interpolation along Street Canyons";
Talk: 2021 IEEE 93rd Vehicular Technology Conference (VTC2021-Spring), Helsinki (virtual); 04-25-2021 - 05-19-2021; in: "Proc. 93rd IEEE Vehicular Technology Conference", (2021).

Providing accurate estimates for Key PerformanceIndicators (KPIs) of mobile cellular networks is crucial to fulfill agreed-upon service requirements. Typically, such performance maps are based on measurements, which are expensive and only offer sparse samples of the underlying ground-truth. Therefore,we require interpolation schemes to further increase the spatial resolution in the area of interest. In dense urban areas, the propagation environment is often governed by tunneling effects along streets and blockages introduced by buildings. However,these aspects are typically not accounted for in state-of-the-art interpolation approaches. As such, we propose a propagation-aware interpolation scheme based on Gaussian Process Regression (GPR), which utilizes available geospatial information via the diffusion kernel. Based on simulations in a stochastic Manhattan Grid (MG) environment, we show that our approach is well equipped to handle environments dominated by tunneling effects and significantly outperforms the omnidirectional Radial Basis Function (RBF) kernel. Further, the proposed graph-based scheme offers a high degree of flexibility in modeling the environment. Therefore, it is not specific to the MG scenario but can easily be extended to generic blockages and real-world street layouts. We show this by applying our approach to a set of Reference Signal Received Power (RSRP) measurements from an operational LTE network collected in a drive-test campaign in Vienna, Austria - again resulting in an error reduction.

Providing accurate estimates for Key PerformanceIndicators (KPIs) of mobile cellular networks is crucial to fulfill agreed-upon service requirements. Typically, such performance maps are based on measurements, which are expensive and only offer sparse samples of the underlying ground-truth. Therefore,we require interpolation schemes to further increase the spatial resolution in the area of interest. In dense urban areas, the propagation environment is often governed by tunneling effects along streets and blockages introduced by buildings. However,these aspects are typically not accounted for in state-of-the-art interpolation approaches. As such, we propose a propagation-aware interpolation scheme based on Gaussian Process Regression (GPR), which utilizes available geospatial information via the diffusion kernel. Based on simulations in a stochastic Manhattan Grid (MG) environment, we show that our approach is well equipped to handle environments dominated by tunneling effects and significantly outperforms the omnidirectional Radial Basis Function (RBF) kernel. Further, the proposed graph-based scheme offers a high degree of flexibility in modeling the environment. Therefore, it is not specific to the MG scenario but can easily be extended to generic blockages and real-world street layouts. We show this by applying our approach to a set of Reference Signal Received Power (RSRP) measurements from an operational LTE network collected in a drive-test campaign in Vienna, Austria - again resulting in an error reduction.

5G, anticipatory networks, diffusion kernel, geospatial, GPR, LTE, performance maps, RSRP, signal strength

http://dx.doi.org/10.1109/VTC2021-Spring51267.2021.9448812