Study On The Method And Application For Urban Road Traffic Signal Regional Eauilibrium Control

Regional traffic signal coordinated control is one of the hotspots in the research field of intelligent traffic management.It aims to estimate the control parameters of each intersection(such as subarea division,cycle,split,offset and phase sequence)by synthetically analyzing the sensor data in the road network.The traffic signal control plays as one of the key element in traffic control,which is still of great research value.To avoid potential risks such as the spread of congestion in the intersection,the anxiety for continuous red lights and the danger of pedestrians while crossing the roads.The problem becomes more complicated and challenging when take coordination of regional road network into consideration,as the traffic flow interaction and dynamic fluctuation between adjacent intersections are quite complex.According to the optimization targets,existing studies of regional coordination control can be divided into green wave optimization method based on time-distance graph and the method of benefit indicators optimization(saturation,stops,delay,travel time and so on)based on traffic flow simulation.While considering about the differences of traffic flow states,it can be divided into the regional coordinated control for unsaturated traffic flow and the regional coordinated control for saturated traffic flow.Basically,the green wave method is more suitable for regional coordinated control for unsaturated traffic flow,while the method of saturation control was used widely in the coordinated control to solve the bottleneck problems.Meanwhile,comparing with the traffic signal control under traditional data acquisition and communication mode,traffic signal control oriented to vehicle infrastructure cooperative has attracted attention and been researched world widely in recent years,which represents the developing tendency in the future.This thesis did research on the urban road traffic network with special heighlight in traffic signal regional coordination control for equilibrium/balance.The research has built adaptive area coordination control models from three aspects,unsaturated traffic,saturated traffic and vehicle infrastructure cooperation.It achieved the equilibrium/balance of area traffic parameters(traffic saturation and green wave bandwidth),and proposed several feasible solutions.The main work of this thesis includes:(1)Considering that the existing study has not taken advantage of fixed-point detection data in accurately judging traffic conditions,a new algorithm was proposed to estimate the traffic state in the intersections based on the Stacked Denoising Autoencoder model.Compared with the traditional method,the algorithmhas non-parametric feature.Furthermore,the algorithm is more consistent with the fuzziness and uncertainty in traffic,and it is more adaptable to estimate the traffic state based on traffic fixed-point detectors data.Experimental results showed that compared to the decision tree model,the overall accuracy of the stacked denoising autoencoder based algorithm obtains certain improvements.While applying the method in traffic state estimation of the road network,it can judge whether the road network is in the saturation state or not,and then provide a criterion for the application of different control methods under unsaturated or saturated road networks.(2)In order to solve the green wave coordination control problem at the road network level under the unsaturated traffic state,this paper proposed a traffic signal network green wave coordination control method to keep the road bandwidth within the specified equilibrium/balance range based on double layer model.The first layer is the strategic layer,where mixed integer linear programming equations were established to haddle the green wave problem of all the arteries in the road network.According to the three principles of distance,cycle and bandwidth,we solved the equations to achieve sub-region division dynamically,with the basic constraints that the green wave bandwidth is balanced.Then,the traffic signal timing parameters in the strategic level of road network can be obtained.The strategic layer algorithm was executed one time every N signal cycles.The second layer is the tactical layer.It took the phase flow as input and the minimum delay as the target function.The split was adjusted based on the reinforcement-learning algorithm and performed one time every signal cycle.Compared with the traditional algorithm,this algorithm can realize bidirectional asymmetrical green wave coordination control of the whole road network in the condition that the road bandwidth was in the equilibrium/balance range and the delay was optimized.Experimental results showed that the proposed method achieved an effective promotion in functioning and accuracy when compared with other mainstream methods,and also achieved a balanced bandwidth.(3)In order to solve the control problems of bottleneck area at road network level under the unsaturated traffic state,a signal control method for traffic bottleneck area which controlling the phase saturation within the specified equilibrium/balance range based on grey reinforcement learning was proposed.Firstly,probability grey numbers were used to represent traffic saturation and traffic service level with interval uncertainty.The goal of 'controlling saturation of each intersection within a certain range' was transformed into that 'controlling the grey state of each intersection close to the optimal value'.Then,the grey reinforcement-learning model was used to search for the optimal traffic signal release scheme,which avoided the derivation of complex function relationship between traffic flow and signal timing schemes.Meanwhile,the neural network was used as the parameter storage model to avoid the so-called"dimension disaster",and the generalization ability of the model was therefore improved.Experimental results showed that the proposed method worked effectively when compared with the traditional pre-timing control method.Furthermore,the method is of strong robustness and achieves the equilibrium/balance saturation.(4)In order to solve the problem of area coordination control of traffic signal in the future environment with the vehicle infrastructure communication,a control method for traffic signal regional coordination and traffic guidance optimization was proposed.In the proposed method,the phase saturation was controlled within a specified range by using gridding and dynamic programming.Firstly,the grid model of road network area was built up to discretize the input and output for parallel computing.Probability grey numbers were used to describe the vehicle position to improve the algorithm accuracy.Then,a dynamic programming model was used to describe the control process,and an iterative updating algorithm was designed for the coordinated optimization of the traffic signal control scheme and the path guidance.Experimental results showed that,when compared with the traditional pre-timing control and actuated control,the proposed method can effectively reduce the average travel time and delay in the road network,and a balanced saturation can thus be achieved.