Efficient management of the resources assigned to maintenance of public road pavements
requires that information about the state of them is updated and in a way that helps the most
appropriate decision in each moment. This process is widely covered for the management of
large road networks, made by surveying with high-performance vehicles.
Main drawback of this process is the cost of this equipment, that is not available to all public
administrations, especially those that manage small local networks, such as councils or town
halls. In these cases, road surveying is carried out in a rudimentary way, usually by visual surveys
filled out by technicians, which leads to a subjective pavement assessment.
To reduce the subjectivity associated with the visual method, SIMEPU project was developed.
Specifically, the project proposes a two-step method consisting of two concatenated
Convolutional Neural Networks (CNNs), first one for distresses identification and next one to
quantify predicted distresses. It also proposes an innovative data collection methodology using
low-cost video cameras with GPS, located at the rear of passenger cars to assess pavement
Furthermore, information ends up being managed by a Geographical Information System (GIS),
with predictive models of machine learning to forecast the state of the network in the future,
adding a multi-objective optimization tool and prioritization of alternatives that determines
actions to be carried out, within an assigned budget and established social and environmental
This methodology allows for a more efficient and reliable pavement assessment, minimizing cost
and time required by the current visual surveys.