Research On Query Algorithm And Lane-Changing Model For Microscopic Traffic Simulation Systems

The urban traffic congestion now becomes a worldwide problem. It is not only inconvenient for residents, but also causes great pollution and the waste of natural resources. Being an important tool in transportation, traffic simulation evaluates and optimizes traffic management plans according to the design and management of roads, traffic flow and traffic lights in a simulation. The simulation results are helpful in improving the utilization efficiency of roads and solving congestion problems. Microscopic traffic simulation is considered as a key technology in analyzing the characteristics of traffic flow. Due to the advantage of simulating behaviors of individual vehicles, microscopic traffic simulation systems are widely researched. Meanwhile, the increasing demand of simulation scale, accuracy and efficiency makes the research on key technologies of microscopic simulation systems a hotspot in recent years.The efficiency problem of nearest neighbor vehicle queries, the adaptive-acceleration problem of safety lane-changing model and the estimation problem of large-scale OD(Origination-destination) matrix based on bus IC card data are main problems in the application of microscopic traffic simulation systems. This dissertation selects key technologies of microscopic traffic simulation systems to be research objects of which main research content includes the efficiency of nearest neighbor vehicle query algorithm, the accuracy of lane-changing(LC) model and the estimation of OD matrix based on bus IC card data. The main contributions of this dissertation are summarized as follows:1. We propose a local-linked nearest neighbor query algorithm to solve the directed multi 1-dimensional NN query problem in the microscopic traffic simulation system. Such an algorithm can be viewed as a variation of B+-tree. It maintains position threads of local vehicles to combine advantages of both B+-tree and linear method. The theoretical analysis and simulation results show that those proposed algorithms can improve the efficiency of microscopic traffic simulations.2. We build an adaptive-acceleration safety lane-changing model to improve the accuracy of microscopic traffic simulation systems. Such model divides the lane-changing process into four stages according to critical conditions of all possible collisions. The model considers the acceleration change in different stages to improve the accuracy of the calculation of safety lane-changing distances.3. We propose a large-scale OD matrix estimation method based on bus IC card data to get the dynamic OD matrix of a city. Such method can estimate the boarding station of each record without GPS and transit scheduling data. The large-scale estimation problem of the boarding station of each record is considered as the graph traverse problem. By using greedy growth algorithm and breadth-first traversal strategy, we reach a local optimum solution to estimate the large-scale tour-based OD matrix.4. We design and implement a light weight prototypal microscopic traffic simulation system to evaluate the performance of proposed algorithm and model. Such a system uses OD matrix based on IC card and real traffic initialization setting to simulate real traffic. Our simulation results show that the NN query algorithm can reduce the time cost of a simulation, and the lane-changing model can provide more accurate safety lane-changing distance.This dissertation studies the practical problem of key technologies of microscopic traffic simulation, and provides an efficient solution. The researching results provide theoretical and technical supports to the study and application of microscopic traffic simulation system. Besides, they can be generalized to a wider range of traffic simulation scenarios.