Research On The Modelling Of Network Traffic Flow Considering Signal Control

Traffic control is one of the most important transportation management measures, characterized by direct linkation with traffic flow. Recent years, with the deterioration of traffic state, control benefit is offset by the unbalance between traffic supply and traffic demand. Control performance of traffic system lies in traffic control theory, the core of which is network traffic model. This dissertation focus on network traffic flow, intend to establish descriptive dynamic traffic flow model and network flow control model.Network traffic dynamics is the integration of basic element dynamics. Dissertation begins from basic element traffic model, namely basic road and signalized intersection. As for basic road, firstly asymmetrical lane changing model is established, which describe objective-oriented and efficiency-oriented lane changing type with a few parameters. Together with LWR theory, basic road model is established. This model can reproduce queue spillover phenomena at channelized section. This model is then used to estimate capacity loss due to channelized section spill back. Basic road is simplified in order to model network model easily later on. As for signalized intersection, feasible phases are selected using graph theory. The selection process is transformed to the refinement of strong unconnected sub-graph of conflict graph drawn from conflict matrix.Proposed model can accomodate to different types of channelization such as left-turn prohibiting, waiting Area for Left-turn and so on. These two models proposed bears some flexibility compared to current models and suit for various scenarios.Oversaturation is directly embodied by long traffic queue and bears complex nature during evolution. Dissertation first determines the critical condition for oversaturation queue. This condition is identical to that derived from demand-supply analysis. Then starting from simplest condition, i.e. one-way link, a novel concept, virtual signal, is proposed to describe queue propagation. This method can describe queue propagation with two parameters:arriving flow and split. Later on complicated conditions are took into account include endogenous flow, channelized section spillover and so on. From the description, we can see that method proposed can not only trace network queue, but also quantitatively describe spillover phenomena include channelized section spillover and link spillover.Dynamic traffic loading models are then established based on methods above. Firstly, road model and route model are established which take road and route as input-output system.These models grasp road dynamics of each cycle by tracing two characteristic vehicles, three characteristic points and four characteristic moments. Based on models above, dynamic network loading model is then established. Compared to current models, our model can adjust various channelization types. Furthermore, because control parameters are explicitly modeled, the method can quickly be used for control optimization. Except that, the model retains the capacity for modeling ITS measures such as route guidance.Based on models refereed to above, three traffic state indexes, i.e. delay, stops and queue length are derived. And then traffic control strategies are proposed for isolated intersection and whole network respectively. With regard to isolated intersection, control strategy that takes channelized section into account with flexible phases derived in Chapter 2 was established. The model transforms the spillover modeling to the calculation of virtual red time. The model can efficiently enhance control performance of isolated intersection.With regard to network control, two control strategies were established based on dynamic network loading model, one is delay minimization and the other is PI minimization. Comparison revealed that PI model can reduce queue about 20%. The reason was that stops term in PI was proportional to queue length.In summary, this dissertation studies signalized network traffic flow. The overall objective is modeling and controlling methods. Research results enhance the comprehension of oversaturation on one hand, and prove the effectiveness of proposed control method on the other. This dissertation bears some value for control algorithm design, realization and application.