Getting smarter urban mobility in Mexico city: visualizing streets pavement conditions and anomalies through fog computing V2I networks and machine learning

Abstract

One of the mayor problems of Mexico City today is its poor mobility. Despite the constant efforts and monetary investments to emerge as one of the largest smart cities in the world, City government has not found a solution to mobility problems. Without a complete coverage of public transportation, and around 35 million daily commuters, with a third of them lasting more than an hour, the country’s capital has been considered several times one of the world's most congested cities. One way to make better decisions about public spending on mobility is to analyze data related to the conditions of its cities’ streets and avenues. Generally, the streets and avenues are fixed as soon as they have a citizen report or when a major incident happens. However, it is uncommon for cities to have real-time reactive systems that detect the different problems they have to fix on the pavement. Today, the government provides internet connection in more than 13,690 Access Points distributed throughout the city. This document thoroughly reviews the context of Mexico City, its communication system range, and technological limitations. At the same time, it is proposed to implement a distributed computing network that couples with the existing infrastructure to capture, filter, and analyze the data that could potentially help decision-makers in terms of public spending on mobility through sensors within vehicles that travel those streets daily and connecting them to a fog-computing architecture on a V2I network. This solution detects main road problems or abnormal conditions in streets and avenues by implementing Machine Learning (ML) algorithms to compare roughness against a flat reference. An equipped vehicle obtained the reference through accelerometry sensors and then sent the data through mid-range communication systems. The present work compares the accuracy and F1 score metrics of a soft-Max Artificial Neural Network (supervised MLA) and a K-Nearest Neighbor (Unsupervised MLA) to select the best option to handle the acquired data, and compares both model’s classification in two different avenues close to ITESM CCM premises.