Telefónica uses data science to systematically identify and locate the unconnected, including them in its networks and operations to sustainably bring connectivity to most of Latin America.
The Internet for All project is Telefónica’s flagship for connecting the unconnected in LATAM. Today are more than 100 million people those who live without a secure Internet connection in the Telefónica area. The reasons are multiple, ranging from geography, population density and socio-economic conditions.
Historically, fixed and mobile networks have been designed to achieve maximum efficiency in dense urban environments. The implementation of these technologies in remote and low-density rural areas is possible but inefficient, which challenges the financial sustainability of the model.
To offer the Internet in these sustainable way it is necessary to increase efficiency by systematically reducing costs, optimizing investment and targeted implementations.
Systematic optimization requires a continuous measurement of financial, operational, technological and organizational data sets.
1. Find the disconnected
The first challenge the team faced was to understand how many are not connected and where. The data set was scarce and incomplete, the census was old and the population was highly mobile. In this case, the team used high-definition satellite imagery at the country level, in addition to neural network models, along with census data such as training.
Implementing visual machine learning algorithms, the model literally counted every house and every settlement on the scale of the country. It was then enriched with reference cross-coverage data from the regulatory source, as well as Telefónica’s proprietary dataset consisting of geolocated data sessions and implementation maps.
The result is a model with a visual representation, which provides a map of population dispersion, with overlapping cover polygons, which allows you to count and locate disconnected populations with good accuracy (95% of the population with less than 3% false positives and less than 240 meters of deviation in the location of the antennas).
2. Optimize transport
Transport networks are the most expensive part of implementing connectivity in remote areas. Optimization of the transport route it has a great impact on the sustainability of a network. This is why the team selected this task as the next challenge to face.
The team started by adding road and infrastructure data to the public sources model and used graph generation to group population settlements. Chart analysis (shortest path, Steiner tree) provided transport routes optimized for population density.
3. AI to optimize network operations
Connecting very remote areas, optimizing operations, and minimizing maintenance and upgrade are key to a sustainable operating model. This line of work is probably the most ambitious for the team. And it is that, when it can take 3 hours by plane and 4 days by boat to reach some locations, being able to be sure that you can detect, or rather, predict if or when you need to perform maintenance on your infrastructure is key.
Just as important is how you design your routes to be as efficient as possible. In this case, the team built a trained neural network with historical failure analysis and powered by network metrics to provide a model capable of monitoring network health in an automated manner, with prediction of possible failures and an optimized maintenance path.
I believe that the kind of data-driven approach to complex problem solving demonstrated in this project is the key to the sustainability of network operators in the future. It’s not just a rural problem, and increased efficiency and optimized deployment and operations are needed to further reduce costs.
