دانلود رایگان مقاله بهینه سازی مشکل زمان بندی نور ترافیک شهری

Abstract

This study addresses urban traffic light scheduling problem (UTLSP). A centralized model is employed to describe the urban traffic light control problem in a scheduling framework. In the proposed model, the concepts of cycles, splits, and offsets are not adopted, making UTLSP fall in the class of model-based optimization problems, where each traffic light is assigned in a real-time manner by the network controller. The objective is to minimize the network-wise total delay time in a given finite horizon. A swarm intelligent algorithm, namely discrete harmony search (DHS), is proposed to solve the UTLSP. In the DHS, a novel new solution generation strategy is proposed to improve the algorithm’s performance. Three local search operators with different structures are proposed based on the feature of UTLSP to improve the performance of DHS in local space. An ensemble of local search methods is proposed to integrate different neighbourhood structures. Extensive computational experiments are carried out using the traffic data from partial traffic network in Singapore. The DHS algorithm with and without local search operators and ensemble is evaluated and tested. The comparisons and discussions verify the effectiveness of DHS algorithms with local search operators and ensemble for solving UTLSP.

5. Conclusions and future works

Inthis study, a centralizedurbantraffic light schedulingproblem (UTLSP) is described. A discrete harmony search (DHS) algorithm has beenproposedtominimize thenetwork-wisedelay time within a fixed time horizon. To improve the DHS algorithm’s performance, a novel new harmony improvising method and a mechanism to limit iteration numbers have been proposed. The improved DHS algorithm aims to balance the exploration and exploitation performances. Three neighbourhood structures and corresponding local search operators have been utilized based on the feature of the UTLSP. Moreover, an ensemble of local search operators has been proposed to integrate three local search operators. In the experiment section, the DHS algorithm has been used to solve sixteen cases generated from real-life traffic data in Singapore. The DHS algorithm and its variants with and without local search operators and ensemble have been compared to the fixed cycle traffic light control system (FCS) to show the effectiveness of the proposed methods. The comparisons and discussions have shown that the HS algorithm as a meta-heuristic has a better performance than the FCS. Indeed, the proposed DHS algorithm could considerably improve the results ofthe standard HS algorithm, and the DHS with ensemble has been proved to achieve the best performance among those achieved by other variants for the UTLSP. As the future studies, we will extend our work in the following directions: 1) to improve the proposed traffic network flow model by considering more real-time constraints and including eventdriven features to well capture drivers’ decision processes – the latter makes it possible to bring the discrete-event system theory in the traffic light scheduling framework with all sorts of advanced modelling techniques as illustrated in, e.g., [45–47]; 2) to explore more efficient scheduling algorithms which are able to tackle both high computational complexity and uncertainties.