A demand-responsive traffic control system for urban areas
Tutor / director / evaluatorBarceló Bugeda, Jaime
Chair / Department / Institute
Universitat Politècnica de Catalunya. Departament d'Estadística i Investigació Operativa
Document typeDoctoral thesis
PublisherUniversitat Politècnica de Catalunya
Rights accessOpen Access
The goal of this Ph.D. Thesis is the design, development and testing of a 'well-engineered system' aimed at the demand-responsive traffic control of urban areas. Because of the author's background -software engineering-, a 'well-engineered system' means a system that is both efficient at performing its task in a wide range of conditions and also that has been built from the robust design's and ease-of-use's points of view. The work has proceed in four steps and, accordingly, this document has been divided in four parts that are briefly desc1ibed here. PART I is, mainly, an introduction to the subject. It introduces the concept of demand responsiveness in traffic control, providing a historical perspective from traditional fixed control systems to current modern demand-responsive systems. The part concludes with a study of areas where research and improvement is needed. All of these areas are considered in the following parts, where some solutions are proposed. PAR T II presents the design, development and preliminary tests of a demand responsive traffic control system, CARS V l, intended to operate in signalized urban areas: networks, arte1ials, and isolated intersections. A graphical user interface, X-Windows and DecWindows compliant, allows the user to specify network characteristics in a friendly and intuitive manner, without the need to be acquainted with the actual modeling. The system features an underlying simulation system and a prediction model based on real-time measured conditions, implements a centralized approach based on small variations, and has flexible detector positioning-and-number requirements. PART III presents the design of a simulation environment, named GETRAM, that solves the difficulties in testing CARS that were discussed at the end of the previous part. We have seen that these difficulties are inherent in the traffic engineer having to use diverse models in order to analyze a traffic network, and that there is the need for a system to salve it. GETRAM provides a unified framework integrating various types of traffic models and tools for traffic analysis, sharing a DataBase, a graphical editor and a mod le for results presentation. The traffic network can be partitioned into views, hierarchically organized polygons in the real world, so that, for example, a simulation model applies only to one of these restricted areas. Network statcs produced by one model can be used as a starting point by another modcl. I n order to ease the task of integrating a new model or analysis tool, a library of object-based high-level functions provide a view-aware access to the DataBase, maintaining consistency. Included in this Ph.D. thesis are the design of the whole environment -DataBase, G ETR AM API, and graphical editor - and the development of the DataBase and GETRAM API. PART IV starts from the demand-responsive traffic control system developed in PART II, CARS V l , and improves it in various respects. First, in order to ease the task of testing the system, it is integrated into the traffic modeling and analysis environment described in PART III. Second, certain parts that directly influence to the effectiveness of the system, such as control timing, adaptive control logics and communication with the controllers, are revised or totally redone. A suite of tests has been applied to the resulting system, CARS V2, in the four scenarios desc1ibed in PART 11. Finally, to further testing the system and taking advantage of the fact of having real-world data available, it is compared against a vehicle-actuated control at an isolated junction.El objetivo de esta tesis es el diseño, desarrollo y test por simulación microscópica de un sistema autoadaptativo apto para cruces aislados, arterias y redes urbanas complejas. El sistema produce planes de control acíclicos y presenta unos requerimientos de tiempo real muy flexibles debidos a utilizar una nueva secuencia cíclica de tareas en la que se predice el estado del sistema a corto término antes de probar planes de control alternativos. Estas pruebas se realizan mediante un modelo interno de simulación que sigue un enfoque mesoscópico a base de paquetes de vehículos de velocidad variable, con el que se consigue modelizar la dinámica de colas de vehículos de forma más exacta que con los sistemas actualmente existentes. Esto da una ventaja, corroborada en los test, en condiciones de flujo altas, con lo que el sistema de controles es capaz de mantener una buena efectividad en un amplio rango de condiciones de tráfico. El sistema viene acompañado de un entorno de simulación y test que aporta un alto grado de integración y de facilidad de uso, a lo largo de todo el proceso de especificación de geometría, parámetros y ejecución de simulación se mantiene una vista de la red de tráfico altamente realista.