Resumen de Data mining, optimization and simulation tools for the design of Intelligent Transportation Systems
María Teresa Sánchez Rico
Over the past years, the problems caused by traffic congestion in urban areas are increasingly severe and visible. Accidents, traffic jams and environmental damage are only some of the negative externalities caused by traffic congestion that today concern society. The so-called Intelligent Transportation Systems (ITS) are multidisciplinary tools which involve the application of advanced technologies and analytical approaches with the purpose of solving some of the referred transport problems. In the last decades different initiatives and projects have emerged in this field. Nevertheless, there is currently no proposal about a generic framework of ITS to be adapted in different traffic contexts. In this thesis, different mathematical and optimization models as well as simulation and data mining approaches are provided with the aim of contributing to the ITS field. The initial motivation for the development of this thesis comes from considering that nowadays any ITS must be capable of fulfilling four main premises: scalability, technological independence, robustness and facing the use of big data. The main contributions of this thesis are: - Literature review. A literature review on the current advances in the field of ITS and its classification is done. Furthermore, a proposal for and ITS prototype for traffic control and management is carried out. - An approach to face the continuous dynamic network loading problem. A continuous DNL model based on flow discretization¿s, instead of time discretization is presented with the main goal of achieving a trade-off between precision and computational cost. - A proposal for the enhancement of traffic signal optimization. A bilevel optimization model based on Time-Of-Day (TOD) intervals for traffic signal timing is proposed to address simultaneously the segmentation problems and the traffic control problems over these time intervals. The model has been solved by the use of efficient metaheuristic algorithms. ¿ An approach for daily traffic patterns identification. A prototype of urban traffic control system based on a prediction-after classification approach is presented. In an off-line phase, a repository of traffic control strategies for a set of (dynamic) traffic patterns is constructed through dynamic cluster techniques. In an on-line phase, the current daily traffic pattern is predicted within the repository and its associated control strategy is implemented in the traffic network. - A proposal for the design of ITS. A proposal for an ITS for traffic control and management is presented by using TOD intervals. This approach replaces the short-term traffic predictions by a finite sequence of stationary states within a set of given traffic patterns.
