The Study On Identification, Dispersion And Simulation Of Urban Traffic Congestion

Urban traffic congestion problems become more and more serious with China’s economic and social development, the increase of people’s income and the acceleration of urbanization process. Traffic congestions bring huge disturbance on daily lives of urban people:the increased economic and time cost for travelling, increased uncertainties, bad driving experience, resource waste of gasoline and air pollutions, etc.A transportation system is a large-scale complicated system affected by human behaviours. It includes the factors from inside, e.g. the distribution of traffic flows and complicated relations among different traffic parameters, to outside ones, e.g. policies, travelling demands, weather conditions and many others. It is ruled not only by objective factors such as road conditions, but also by subjective behaviours such as travelling needs and driving habits. It is a mixture of observable variables such as traffic flow, speed and density, and many unmeasurable variables such as route selections and real-time proportion of specified vehicle types.The transportation system is nonlinear and highly uncertain, where a tiny parametric fluctuation may bring a completely different consequence, which is often called "butterfly effect". Therefore, deterministic methods are not sufficient in studying certain phenomena, where probabilistic analysis takes their places. The system is also open to the outside world that has not been modelled yet. The correlations and interactions are everywhere, between different parts of the system, and between the ones inside and outside the system. New models are required to be continuously integrated into a transportation simulation platform, and traditional closed simulation architecture is not suitable for the scenarios.Focused on the above features of a transportation system, the paper uses the technology to deal with uncertainties in the analysis of spatio-temporal distribution of transportation congestion. The paper also uses a hybrid method from trajectory-based statistical methods and model-based simulation methods to study congestion-releasing controls. Besides that, an innovative transportation simulation platform is developed by the author, in order to implement simulation-based transportation analysis considering uncertain factors and to extend models in case of need. For the above purposes, the following works have been done:1. Unstationary traffic parameter trajectories may stop the efforts to increase their forecast precision and achieve reasonable correlation analysis. In this paper,2 decomposition methods for a time-variant transportation parametric curve into trend and detail components are presented to solve the problem in different scenarios. One is the wavelet analysis used for the decomposition of a single curve. The other is the polynomial fit function and the smart segmentation algorithm, which divides the correlated signals according to their correlation features before decomposition. These methods improve the stationarity of the trajectories, avoiding the interference between signals in different time scales, and realizing the multi-scale analysis on transportation signals.2. Based on stationary trend component of a transportation signal, the forecast result of the signal’s future trend is obtained by using the ARMA algorithm. Considering the effect of uncertain factors on the trend, the Monte-Carlo method is used to obtain random combinations of upward or downward detail waves in the trend curve’s unstable region. A possible trend set is created based on the scenarios under different combinations, which leads to the emergence of interval forecast result. A comparison between the measured, the deterministic and the interval trend forecast results shows the effectiveness of the proposed method.3. A trajectory-based index, and a probabilistic index for transportation congestion judgment are proposed. The patio-temporal distribution features of traffic flow are discussed, and the extended correlation indices, as well as related algorithms, are invented. Based on the new tools, traffic flow’s spatio-temporal correlations can be analysed in detail. The congestion judgment method considering the correlations is also presented, which is useful in congestion pre-warning for neighboring roads. As a region-level tool, the clustering index of intra-regional correlation is presented, in order to make clustering analysis and reduce the scale of solving problems in case of studying traffic controls for complicated transportation networks.4. An iterative searching algorithm is created to estimate the parameters of the dynamic transportation distribution model on multiple regions divided by the clustering index of intra-regional correlation. The error between simulated and measured results of every region’s leading road can be reduced when the dynamic O-D parameters are tuned step-by-step. The iterative searching result is a dynamic transportation distribution model which well fits the measured parameters.5. A complete framework is presented to build optimized congestion-releasing guidance strategies considering the uncertainties of forecast parameters. Under the framework, the objective function for traffic guidance is established to obtain system optimum. The open-loop traffic simulation model is then built and uses the probabilistic forecast results as its boundaries. These boundaries will constrain the variance of simulation parameters. A smart optimization algorithm is invented based on the sensitivity (the variance of objective function value divided by the guidance strategy’s disturbance). Therefore, the variance of road speed can be induced, and the possible solution will be eventurally obtained if it passes the filtering by using the thresholds of congestion probability. Case studies indicate that the result automatically obtained by the algorithm well fits the actual experience on removing traffic congestion.6. The Dynamic Simulation Platform for Traffic Congestions (DSPTC) is built to be a dynamic, interactive, open and scalable platform. Technical achievements about analysis, judgement, forecasting and guidance optimization of traffic congestion in this paper are all integrated together in this platform in the form of simulation components. The simulation of traffic congestion and guidance result is therefore realized. DSPTC also integrates GIS and database systems, and provide storage, data mining, simulation, calculating and intrusive display functions for different traffic data from collection, analysis and decisions.The above work selects valuable technical aspects as breakthrough points, and passes through 4 vital steps in transportation congestion studies for urban networks:theoretical analysis, algorithm creation, control optimization and software applications. The mechanism analysis, state estimation, forecasting and pre-warning, and optimization for potential traffic guidance measures surrounding the congestion problems are all involved and new advancement is achieved in above research. Besides that, a simulation platform is newly developed for industry use, which applies advanced IT technologies and provides solid infrastructure for future deep applications.