Previsão em tempo real de condições de tráfego em redes veiculares

• Autor(a) principal: Jorge Miguel Marques dos Reis
• Data de Publicação: 2016
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://repositorio-aberto.up.pt/handle/10216/86699
• topic: Engenharia electrotécnica, electrónica e informática; Electrical engineering, Electronic engineering, Information engineering

One of the biggest challenges proposed nowadays to Intelligent Transportation Systems (ITS) is the support to the reduction of traffic congestions due to the high flux of vehicles that travel daily to the major metropolitan areas. The traffic management system NEXT, developed by Armis ITS company, currently allows real-time forecasting of traffic conditions on road networks. Its implementation is based on the use of simulation techniques (model-driven approach) and data mining (data-driven approach), using historical data collected from road sensors.However, the different forecasting techniques currently in use generate different results. There is a module that calculates the efficiency (percentage of accuracy) of each technique, considering the error between prediction and reality. Choosing the appropriate technique mainly depends on the system's ability to identify the situation and context where a given technique is distinguishable from the others. This context is closely related to the state of the network, which can be characterized by various metrics that influence the flux of vehicles, such as date and time, weather conditions, the occurrence of incidents or the network topology. So, be aware of the relationship between the state of the network and the suitable technique is particularly imposed by the need to make predictions in real time.This way, the main problem to be studied is the efficient identification of the best forecasting technique to apply in certain cases and situations.The solution is to study, analyze and implement an intelligent module based on machine learning techniques, to choose the best forecasting algorithm, between the various algorithms currently implemented in NEXT, considering its accuracy for each state of the network.As a result, it is expected to improve the efficiency of prediction algorithms, which can have a big impact on the management and control of road traffic, and can be applyed in real time information systems.The project development will be guided to full integration into NEXT system, and may be available in traffic control centers of Infraestruturas de Portugal (IP).