Artificial Intelligence applied to the determination of the valid travel routes of Nova Ferroeste

A 1,304-km long Brazilian interstate railroad that connects parts of the states of Mato Grosso do Sul and Paraná.

Huge quantities of data, complex topography and a need to integrate the line into existing terminals and rail networks saw Artificial Intelligence (AI) adopted, with  Machine Learning applied to determine the optimal options for the railroad route.

To do this, a multi-criteria study was carried out involving 35 variables such as market, physical environment, logistics, socio-environmental and socio-economic characteristics.

The team first had to collect and process data using an algorithm to establish what was true and useful for the project. Several processes were automated using the Python and ArcPy languages. The data was then converted to raster (image), normalized in 8 bits, combined and distributed in five information layers: the physical, logistical, market, socio-environmental, and socioeconomic dimensions.

The categorised data was used to generate maps showing the regions of greatest and least favourability according to all indicators. The team then developed a Machine Learning model to establish corridors with the lowest cumulative cost, and to identify the best areas for Nova Ferroeste, taking into account building, environmental and operational aspects.

As an innovation, the impact has been profound, opening up new methods for identifying optimal routes on the basis of various physical, marketing, socioeconomic and socio-environmental outcomes, including reduction of noise and traffic jams, well-to-tank emissions, damage to native habitats, accidents, pollutant emissions and transport costs. It also allowed for optimisations of increases in job and income generation, tax collection, real estate valuation and climate bond certification.