Forecasting mobile transmission reliability using crowd-sourced cellular coverage data

Abstract

As smartphones work their way into mission-critical applications, there is a need to gain knowledge of access network speeds and their variability at different locations. This information is vital to ensuring the efficient transmission of time-sensitive data and can mean the difference between life and death for some patients [1]. However, the decision to adopt smartphone applications (apps) that provide advanced communication services is complicated by the uncertainty of whether they will actually perform well in the field. Reliable operation in the field is hard to assess from laboratory demonstrations with reliable network coverage. Even if performance can be related to simple measures of network connectivity (e.g. the number of "bars"), objective 3rd-party assessments are difficult to obtain short of extensive field testing. This paper presents preliminary work toward an empirical model that predicts the number of bars in specific geographical locations using "crowd-sourced" signal strength data. Preliminary field test data was used to illustrate a Geographical Information System (GIS)-based end-to-end process by which signal strength (number of bars) in rural areas can be predicted from available signal strength measurements on major thorough fairs. Two linear models for predicting signal strength were developed using predictor variables that are easily assessed using standard GIS software. While the results presented here fail to achieve a high degree of statistical significance, the basic feasibility of the approach is established, and the factors that contribute to success or failure of the approach are discussed.