Please use this identifier to cite or link to this item:
|Title:||Region-based urban traffic management||Authors:||Lentzakis, Antonios||Keywords:||DRNTU::Engineering::Electrical and electronic engineering||Issue Date:||2018||Source:||Lentzakis, A. (2018). Region-based urban traffic management. Doctoral thesis, Nanyang Technological University, Singapore.||Abstract:||Quality of life in cities, especially for vehicle owners, is closely correlated with the delay experienced by drivers, a direct consequence of congestion. Costs associated with delays incurred to travelers during peak periods include the loss of man-hours, increased vehicle emissions, increased fuel consumption and more. The magnitude of the socio-economic impact of congestion is usually very great. Unfortunately, due to limited space availability in modern urban regions, infrastructure upgrade and extension costs are prohibitive, with optimal utilization of available infrastructure being the only viable alternative. This work describes different aspects of a proposed Region-based Urban Traffic Simulation and Management Framework, which takes advantage of the Urban-scale Macroscopic Fundamental Diagram concept. Concretely, weighted feature extensions of center-based clustering methods are introduced and compared with an implementation of normalized spectral clustering. A new, feature weighted extension of k-harmonic means clustering outperforms all other methods when applied to an urban traffic network encapsulating Singapore's Central Business District. The partitioning of the urban traffic network into homogeneously congested regions is an essential requirement for the existence and validity of Urban-scale Macroscopic Fundamental Diagram. Subsequently, a multi-class extension of a regional dynamic traffic model, the Network Transmission Model, is proposed and subsequently used to simulate regional traffic dynamics. Two innovative region-based prescriptive route guidance approaches are proposed, i.e., a non-predictive strategic learning-based routing method and a predictive simulation-based routing method. They are applied on a 16-region, diamond-shaped grid network which makes use of the Network Transmission Model to simulate the regional traffic dynamics. Their performance is compared to multinomial logit routing, a realistic reference case which models self-interested traveler route choice behavior. It should be noted that, all routing methods come integrated with a Public Transit Diversion mechanism. As expected, the predictive routing method outperforms all other routing methods. A market penetration scheme is chosen to be applied to the new multi-class extension of the Network Transmission Model, where three traveler classes are defined. The 1st traveler class makes use of autonomous vehicles, the 2nd traveler class makes use of conventional vehicles equipped with Route Guidance and Information System and the 3rd traveler class makes use of unequipped conventional vehicles. The 1st class is represented by the predictive routing method with full compliance and under two information provision scenarios S1 (imperfect information),S2 (perfect information), that affect travel time prediction accuracy. The 2nd class is represented by the non-predictive routing method with the possibility of non-compliance. The 3rd class is represented by multinomial logit routing, modeling route choice under imperfect travel time perception. Regional traffic dynamics for simultaneous application of all aforementioned routing methods are simulated, for various combinations of traveler class market penetration rates and non-compliance rates. An analysis of the overall network performance gains, as well as individual performance metric results, was conducted for each set of market penetration rate combinations with promising results about the potential impact that autonomous vehicles will have on the overall network performance, as well as the individual benefits provided to all other traveler classes.||URI:||http://hdl.handle.net/10356/75886||Fulltext Permission:||open||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Theses|
Page view(s) 20129
checked on Oct 26, 2020
checked on Oct 26, 2020
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.