TY - THES AU - Cruz Piris, Luis de la AU - Iván Marsá Maestre (dir. tes.) AU - Miguel Angel López Carmona (codir. tes) AU - Sascha Ossowski (pres.) AU - Juan Ramón Velasco Pérez (secr.) AU - Mario Vega Barbas (voc.) T1 - Contribución a las estrategias de optimización multiobjetivo para la coordinación de vehículos en intersecciones urbanas LA - spa PY - 2019/11/14/ PB - Universidad de Alcalá AB - More than 72 % of Europeans live in urban areas. Some estimations quantify the possible cost related to the time lost in traffic jams in the European Union between 2016 and 2026 in more than 200 thousand million euros. Traffic jams are the most common urban traffic problem but not the only one. The environmental impact produced directly by vehicular traffic, both by pollutant emissions and by the noise and discomfort endured by citizens, has increased the number of initiatives to alleviate these problems in recent years. Each vehicle that circulates on a road network can be considered as an independent element that, to achieve its objectives, uses the resources of the network for a specific period. When two or more vehicles need to use a resource during the same time, and their trajectories are incompatible (the passage of one of the vehicles causes the stopping of the other), then the use of coordination techniques is essential. The evolution of technologies related to Intelligent Transportation Systems such as improved sensor accuracy, more powerful processing systems and wireless communication networks that have reduced their latency to levels close to wired networks, has paved the way to apply new vehicle coordination techniques on key points of traffic networks such as intersections. This thesis addresses the problem of the coordination of vehicles as they cross junctions or intersections in an urban environment. The study of this problem begins by analyzing the possible sources of information existing in an urban traffic scenario, considering the technologies currently available. In this area, our first contribution is the proposal of a methodology based on the study of the centrality of the road network. The goal of such methodology is to determine the best positions for sensors, so that they provide the most relevant data for traffic modeling. The central hypothesis of this thesis is that it is possible to improve the optimization processes used for the current management of intersections. To achieve this goal, the first objective is to define a model capable of labeling unequivocally all the elements that make up an intersection. This model is used later as the basis for subsequent optimization proposals. From the point of view of the coordination of vehicles, the contributions of this thesis are classified depending on whether the coordination of vehicles is carried out through external elements such as traffic lights, or this coordination occurs in a scenario of autonomous vehicles, where each of them is able to cross the intersection, without stopping, in safe conditions. In the first case, the proposal focuses on the development of a multi-agent system, deployed on a simulated scenario based on real data, capable of managing the traffic light phases of each intersection. The controlling agents for each intersection can vary their behavior based on the traffic indicators they receive. In the second case, the research focuses on the behavior of vehicles inside the intersections, and the possible methods to generate vehicle arrival patterns that allow their crossing, in safe conditions, without the vehicles being hindered and complying with system flow preferences. To this end, in a first iteration, an optimization process based on a genetic algorithm with fixed paths between the entry and exit points of the intersection has been designed. The algorithm is focused on obtaining the highest possible yield if all entrance flows are equal. Subsequently, this proposal is extended allowing all possible paths between the entry and exit points of the intersection and generalizing the optimization goal to achieve the desired input flows, without the restriction of having symmetrical input flows. Due to the exponential increase in the complexity of the problem in this last scenario, a genetic algorithm with variable-length chromosomes is proposed, adapted for this problem. Simulators and realistic traffic scenarios have been used to validate the proposals, which allow intensive testing of each of them. In addition to the use of commercial simulators, it has been necessary to implement our own intersection simulator, which can reproduce the peculiarities of the problem. In each case, the results obtained using the proposals have been compared with other widely used solutions. The good results obtained with the methods proposed in this thesis allow us to confirm the hypotheses raised at the beginning of the research. L3 - https://aplicaciones.ciencia.gob.es/teseo/#/tesis/O151052/detalle UR - https://portalcientifico.uah.es/documentos/6142af5a27af2147d144509c DP - Dialnet - Portal de la Investigación ER -