Research On Iterative Learning Control Methods For Urban Traffic Signals

The limited development space of urban road networks and the growing vehicles are increasingly aggravating the contradiction between traffic supplies and traffic demands.The resulting consequences are the frequent traffic congestion,which in turn triggers such as environment pollution,energy waste,traffic safety and a series of traffic problems.Practice has proved that these issues cannot be fundamentally solved through the expansion of roads and other infrastructure,so the scientific and effective traffic signal control strategy has become an important measure to solve the problems of urban traffic.A reasonable signal timing plays an important role in improving the traffic flow conditions,alleviating the traffic congestion and improving the capacity of existing road networks.Themed “Research on iterative learning control methods for urban traffic signals”,the paper studies the applications of iterative learning control method in urban traffic signals in many different situations,such as the limited inputs as well as the limited states,uncertain initial states and random external disturbances.Based on the macroscopic fundamental diagrams of road networks,the impacts of the iterative learning based urban traffic signal control strategies on the macroscopic traffic situations of the network are analyzed.The main contributions are summarized as follows:1.In view of the limited conditions of green times for each phase at intersections and the constraints of vehicle queue lengths in each road,a constrained iterative learning based urban traffic signal control method is proposed.The convergences of the iterative learning control algorithms with restricted inputs and constrainted states are analyzed.Through iterative learning control of the intersection signals,the vehicle queues of each road in the network can gradually achieve the equilibrium states.Thus,the traffic congestion probably caused by the long queues of one phase is effectively prevented.The results of a case study with Xiaozhai intersection in Xi'an conducted by VISSIM simulation verify the effectiveness of the proposed method.2.With consideration of the same initial reset condition hardly being met in the iteration process of transportation system and the problem of random external disturbances,an iterative learning control method with uncertain initial state for urban traffic signals is proposed.Also,the convergence of the iterative learning control algorithm with random disturbances is analyzed.Analysis results show that the system tracking errors can converge to a given bounds when the initial states of traffic flow fluctuate around the desired ones in a small range or the system with bounded external disturbances.The bounds mainly depend on the uncertainty of system and the random external disturbances.Finally,with the VISSIM software,the simulation tests with the artery of Zhuque Road in Xi'an are carried out,and the impacts of uncertain initial states and random disturbances on the performance of iterative learning based urban traffic signal control method are analyzed.3.Considering the problems of the poor anti-interference ability and the incapability of controlling the randomly changing traffic conditions in open loop iterative learning control algorithm,an open-closed-loop PD type iterative learning control algorithm is put forward.Open-closed-loop iterative learning control algorithm can calculate the green times by using both the previous and current running information of the system.Therefore,it can effectively restrain the uncertainty and nonlinearity of the traffic system,improve the anti-interference ability of the control system and enhance the system robustness.A case study with the local traffic network composed by four intersections around Xiaozhai intersection in Xi'an is conducted by the VISSIM simulation software.The results show that the open-closed-loop iterative learning control algorithm can achieve better control performance than the simple open loop iterative learning control algorithm.4.In view of the need of a prespecified green time plan and the incapability of long-term adaptation to the randomly changing traffic conditions in traffic-responsive urban control(TUC)strategy,a hybrid urban traffic signal control method is proposed based on the iterative learning control and the TUC strategy.By replacing the prespecified green time plan with iterative learning control in TUC strategy,the hybrid strategy can realize feedforward control of the traffic signals and implement real-time feedback control of the randomly changing traffic conditions.In this way,the control performance of TUC strategy can be effectively improved.With the VISSIM simulation software,a case study with the regional traffic network of Xiaozhai commercial circle in Xi'an is carried out.The results show that the control performance of the hybrid traffic signal control strategy based on iterative learning control and TUC is better than that of the pure TUC strategy.5.Based on the macroscopic fundamental diagram(MFD)model of road networks,the impacts of the iterative learning based traffic signal control strategies on the traffic situations of road networks are studied under different traffic demands.The results of a case study with the regional traffic network of Xiaozhai commercial circle in Xi'an show that the vehicle queues of each road in the network can gradually achieve the equilibrium states through iterative control of the intersection signals,which makes the vehicle density distribution in the network be more homogenous and ensures traffic flows run under a well defined MFD.Therefore,the traffic efficiency decline and traffic congestion caused by heterogeneous distribution of vehicle density are effectively prevented.