Perimeter Control Based On Dynamic Partitioning Algorithm For Urban Road Networks

With rapid development of the national economy and socialization process,the scale of road traffic in Chinese cities is expanding and the trip demand is increasing.Traffic congestion,the critical problem that influences the development of Chinese big cities,has not been well addressed.It brings great difficulties for the cities to manage and control large-scale urban road networks,due to complexity of urban road traffic network,timevariance and uncertainties of the traffic flow in the network.In this thesis,evolution of traffic flow is revealed and analyzed for the large-scale road networks from a macroscopic level.By artificial intelligence technology,road network is partitioned with consideration of spatiotemporal similarity.On this basis,the perimeter control and path induction strategy for macroscopic urban road traffic flow is studied.It is of great practical significance to alleviate traffic congestion and improve the operational efficiency of the urban road networks.This thesis mainly focuses on two research topics,including dynamic partitioning algorithm and switching perimeter control for large-scale heterogeneous urban road networks from a macroscopic level.The contents are summarized as follows。First,a static road networks partitioning algorithm is proposed and extended from normalized cut(NCut)algorithm,by considering the spatiotemporal similarity characteristics of traffic flow time series and introducing the spatial connectivity constraints among sub-regions.On this basis,noticing the dynamic characteristics of the time-varying traffic conditions in the sub-regions of the road networks during the peak period,optimal cluster numbers of sub-regions over different periods are determined by designing an appropriate evaluation index and a NCut-based dynamic partitioning algorithm using traffic flow time series is proposed for the road network.Second,the effectiveness of static partitioning algorithm and dynamic partitioning algorithm are verified by utilizing the speed time series data collected within a road network in the Northeast Second Ring of Beijing.The numerical results show that the traffic conditions in different sub-regions of the network can be identified effectively by the proposed partitioning algorithms.Comparing with static partitioning algorithm,the spatiotemporal evolution characteristics of the traffic flow is fully utilized by dynamic partitioning algorithm,and the performance of road networks partitioning is also improved.Third,due to the size and the boundaries of sub-regions could be changed under dynamic partitioning algorithm for road network,the macroscopic fundamental diagram(MFD)describing the road network’s properties may alter accordingly.On the basis of dynamic partitioning,a switching MFDs model of a two-region urban network is built following the idea of the switching system,and a switching model prediction control method is designed for perimeter control.Finally,by introducing a path induction strategy,a multi-region switching system for a large-scale urban road network is considered,and a switching model predictive perimeter control method with path induction is proposed.The effectiveness and superiority of the proposed perimeter control algorithms is demonstrated through numerical experiments and comparative studies for multi-region urban network.