Dynamic Real-time Feedback Control Of Signal Lights In Urban Saturated Road Network

With rapidly increasing preservation of vehicle,traffic congestion is increasingly serious,especially oversaturated state in rush hours.Nowadays,most of traffic signal control systems are designed for the under-saturated intersection.The control performance of the systems largely decreases when the intersection is oversaturated.Thus,the problem of traffic signal control for the oversaturated situation is still a difficult problem in traffic fields.Then,this paper mainly focuses on the research about traffic signal control under the oversaturated situation.First,the signal-intersection is studied which considers phase switching problem of four-phase intersections.By introducing positive switched theory,the discrete model of queuing vehicles at the oversaturated intersection is proposed.Based on the energy theory,the intersection is taken as the energy storage system and the inflows and outflows of the vehicle are taken as storage and dissipation of energy respectively.According to the energy dissipation,we design the state feedback controller of the effective green time to realize the system non-accumulative,namely to minimize the queue length.We gives the simulation of single-intersection signal control algorithm,and gives the comparison between the algorithm we proposed and the oversaturated signal control algorithm literature existed.The experimental results show that the algorithm we proposed needs fewer cycles to dissipate the queued vehicles.Then,regional traffic signal control is studied.Considering the dynamic characteristics of input flow,the improved store-forward model is proposed.Due to the complexity and coordination of regional traffic,a large-scale regional traffic is divided into several sub-regions using a region decomposition method firstly.Secondly,the optimization problem of the large-scale regional traffic is presented based on Hico-MPC.Lagrange dual theory multipliers are introduced to deal with interconnecting constraints among sub-regions.Then we forecast the vehicle queue of sub-regions and optimize the green light time.Finally,the simulation is presented to illustrate the effective of the improved model by comparing with the store-forward model.The results show that the improved model requires less computation time and low computational cost when the same control effect is achieved.