Algoritmo inteligente para geração de rotas em smart cities

• Autor(a) principal: MORAIS, Renê Douglas Nobre de
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFRPE
• Texto Completo: http://www.tede2.ufrpe.br:8080/tede2/handle/tede2/7856
• topic: Cidade inteligente; Geração de rota; Tráfego de veículo; Computação evolucionária; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

The traffic congestion in large urban centers is one of the main problems for people who need to drive daily. The concept of smart cities has made possible the development of innumerable innovative solutions that, through technology, share and disseminate various information in real time to the population. However, these problems have numerous variables that make it difficult to resolve them in satisfactory time. In this way, this work proposes an algorithm that uses Points of Reference to determine rides in large-scale road networks considering 3 objectives (travel time, distance and number of traffic lights) in a satisfactory ti using an evolutionary multiobjective algorithm. In addition, a new mutation operator that guarantees only minor changes in individuals is proposed. The proposed approach is tested by performing simulations on a map of a large city with different levels of traffic congestion using Openstreetmap data. Finally, the results of the simulations are compared with solutions generated by the Dijkstra’s algorithm that represent the lowest theoretical value for each objective. Five sets of simulations were carried out with 3 scenarios each with different levels of traffic congestion. The results showed that the algorithm found viable trade-offs, highlighting the best results for the highest levels of congestion with average losses that did not exceed 20% in 2 of the 3 objectives in most of the simulations when compared with the theoretical minimum values. As for the computational time spent, in 2 groups the algorithm took on average 3 seconds to find the best routes and 6 seconds on average for the other 3 groups. In this way, it was concluded that the algorithm was able to generate viable trade-offs in a good computational time considering the environment in which it was executed. These results also reflected the efficiency of the proposed mutation operator.