Abstract:
Most of European cities face increasing problems caused by excessive traffic of conventional fuel-based transport modes. To reverse this situation, sustainable urban mobility policies have been promoting soft modes of transport, such as walking. Despite the advantages of walking in reducing traffic congestion and pollution, cities have not always evolved to accommodate the needs of pedestri-ans. According to the European Commission, in 2020, 20% of road fatalities in the European Union (EU) and 21% in Portugal were pedestrian. Pedestrian fatal-ity rates per million population was 9.7 for all EU countries and 13.1 for Portugal. In European and Portuguese urban areas, 36% and 27% of the fatalities were pedestrians’ and 49% and 56% of all pedestrian fatalities were elderly’s (respec-tively). In pedestrian infrastructures, crossings are considered the most critical element due to conflicts between vehicles and pedestrians. It is then essential to identify and minimize risk factors that increase the probability of accidents in these locations. The proposed work intends to assess this challenge by using Ar-tificial Neural Network (ANN) to create pedestrian severity prediction models and identify road and pedestrian risk factors for accident occurred in or near ur-ban crossings. The official Portuguese database on run over pedestrian accidents occurred between 2017-2021 was analyzed with ANN considering two scenarios: pre-Covid-19 and during Covid-19 period. Results obtained demonstrate that the use of ANN can promote a proactive infrastructure management, suggesting that crossings traffic lights operation, lighting, shoulders and pavement conditions, high speed limits (51-90 km/h) and pedestrians moving in soft modes are critical factors.
