| From the point of view of system control and optimization, the urban traffic system is a dynamic system with high complexity. On the one hand, the traffic system has hybrid dynamic property of discrete event and continuous time, high nonlinearity, non-stationary randomness with unknown distribution, fluctuating system parameters according to environmental conditions and people's travel demand, and strong-coupling adjacent intersections. Therefore, the states of a traffic system are difficult to measure, predict or control. On the other hand, several kinds of control action are taken on the urban traffic system, such as signals at intersections, guiding information and etc. Signals are the most commonly used control action. However, signals take effect by the switch of signal phases, which introduces integer control values into the traffic control problem, so the traffic control problem becomes a large-scale mixed integer programming problem, the computation of which requires an exponential-complexity optimization algorithm. In addition, considering traffic security, operational efficiency and people's travel habits, the setting of signal phases must satisfy strong constraints, such as maximum and minimum green time, maximum queue length and smooth switch of signal timing parameters.Although the effectiveness of a model-based traffic control strategy depends on the precision of the traffic prediction model, it's still necessary for such a complex and large-scale system as a traffic system to evaluate variant control strategies with a global prediction model and obtain the global optimal control strategy. There are three important problems within the construction of a model-based urban traffic adaptive coordinated control system (UTACCS). The first is to establish an appropriate mathematic formulation, including a traffic prediction model, an objective function and control variables. The second is to select or design an effective optimization algorithm to perform on-line calculation of the corresponding control problem. The third is to make sufficient experimental verification of UTACCS before field tests and deployment are taken. At present, traffic micro-simulation systems are the best platform to do the test. This thesis deeply investigates the three fields. Two kinds of model-based UTACCS, coordinated optimization of timing parameters (COTP) and rolling-horizon optimization of signal phases (RHOSP), are established. The construction graph of the ant colony optimization (ACO) algorithm is revised to solve traffic control problems. And the ACO algorithm is improved further to obtain better search efficiency in large-scale traffic control problems. Finally, on the Simulation and Analysis System for Urban Mixed Traffic (SASUMT) that was developed by Zhejiang University, the two UTACCSs are compared and analyzed in the cases of an isolated intersection, an arterial road and a traffic region. The detailed research work done in this thesis is listed below.1.The ACO algorithm is a heuristic search method with a solution construction graph that explicitly describes the whole solution space of optimization problems. Therefore, the definition of the construction graph directly influences computation cost and search efficiency. According to the difference of defining solution building blocks and mapping them to the nodes of the construction graph, three kinds of construction graph, simple construction graph (SCG), basic layered construction graph (BLCG) and compound layered construction graph (CLCG), are defined here. The three kinds of construction graph can be applied to optimization problems with different characteristics, and the two kinds of layered construction graph are more suitable for complex multi-stage dynamic decision problems (CMSDDP) than SCG. The CLCG defines smaller solution building blocks and is able to perform better in large-scale CMSDDPs than the BLCG.2.The construction graph of ACO algorithms need statically describe the whole solution space (or discrete solution spa...
|
PDF Full Text Request