Characteristic Analysis Of Inefficient Urban Road Network Driven By Computational Intelligence

The low efficiency of urban road networks refers to areas/ roadways/ intersections with unsatisfied-efficiency are involved in the network,which represents itself as lower travel speeds,extended travel times or longer delays.This problem imposes an adverse impact on society and the economy.Therefore,inefficiency abatement of the urban road network has received increased attention from traffic managers.As an open giant system,the urban road network possesses features of complex,dynamic,and temporal.These inherent features hinder managers from recognizing the critical elements that have an essential impact on the efficiency of the road network or understanding the factors that cause the unsatisfactory performance of roadways and intersections.These confusions significantly undermine the pertinence and effectiveness of the work to govern road network inefficiency.Although the intelligent transportation system has provided abundant data resources and computing capacity for solving these problems,existing methods still see limitations to their adaptability,computing speed,and interpretability when dealing with complex urban road networks.Computational intelligence(CI),a branch of artificial intelligence(AI),simulates human abilities to allow computers more intelligence to handle complexities.It shows superiority in traffic analysis.Given this,with CI and data integrated and mechanism models from multiple disciplines supplemented,this paper proposes a range of methods to identify critical areas,roadways,and intersections based on their impact on road network efficiency.The mechanisms leading to inefficient performance are analyzed at macro and micro levels,which may shed light on improving urban traffic efficiencies.Regarding identifying critical areas that affect road network efficiency and analyzing their features when road network performance level(NPL)changes(in particular,deteriorates),a set of solutions is proposed to address the two issues.This paper builds a deep learning model to adapt to the complexity of road networks and introduces the visualization algorithm to improve the model’s interpretability.To render the model the capacity to evaluate NPL automatically and the profound knowledge about the correlation between elements,a graphical representation of the total amount of road network data and the pre-trained network are involved.Furthermore,the visualization algorithm is deployed to explain the decision-making mechanism of the model,thus revealing the areas that play an essential role under different NPLs.Based on the visualization time series,the transfer of the critical areas under various NPLs is studied in depth.The set of methods improves existing research on road network areas with more consideration of efficiency features.In terms of identifying critical roadways based on their impact on road network efficiency,this paper investigates the defects of existing methods and presents a model to determine critical roadways under their inspirations.As for the “data-exclusion method” referring to road network vulnerability,this paper illustrates its infeasibility with theoretical analysis and data experiments.From the standpoint of roadway function,this paper suggests a data-driven method to determine the essential roadways according to their impact on road network efficiency.The method is proven reliable but flawed in efficiency.Inspired by the above work,a feature set is extracted,which consists of dynamic/static features that may affect the roadway’s importance.With these features,a binomial logit model is constructed to label critical roadways actively.The model holds the advantages of the above algorithms and enhances their efficiency to fulfill the online task of identifying critical roadways.As for the characteristics concerning roadway efficiency,this paper takes advantage of data-driven and mechanism models from multiple disciplines to describe how efficiency indexes are affected by emergencies and macro/micro traffic features,which could help improve road network performance.This paper references several mechanism models in the financial analysis field to analyze travel time fluctuation and develops an optimized algorithm to describe daily events’ influence on travel time.The optimized algorithm functions better than existing models in revealing events’ effect on road efficiencies.Taking the headway distribution as the indicator,mixed distribution models and iterative algorithms are applied to profile the impact of traffic composition,or the large vehicle occupancy,on roadway efficiency.Based on the traffic wave theory and unsupervised learning methods,this paper classifies the overtaking behavior on urban roadways and displays the mechanism of its impact on roadway performance.All above methods have been implemented on large-scale data sets,which protects the results from errors caused by simulation or small data sets.In addition,microscopic characteristics of roadways are fully concerned,rendering the methods good adaptability to the fuzziness of traffic.When it comes to identifying critical intersections based on their impact on road network performance,this paper suggests a fuzzy temporal network model concerning temporal and spatial correlations between intersections.Drawing on the temporal network model and fuzzy theory,the model characterizes road network’s temporality,dynamics,and randomness.The kernel of the model,the fuzzy supra-adjacency matrix,presents the importance rank of all intersections in studied intervals.The model takes care of the temporal and spatial correlations between elements and the impact duration/strength of intersections on their surrounding areas.Therefore,compared with the saturation model and Page Rank series model,the proposed model is more consistent with the characteristics of the road network.Towards efficiency characteristics of intersections,this paper conducts research with the license plate recognition data.Taking headways and delays as indexes,this paper investigates the factors contributing to intersection inefficiency.A dilemma is observed that it is difficult to synchronize the passing vehicle data with the signal control scheme for the same intersection.To solve this,this paper proposes a method based on the credibility methods to determine all parameters of the intersection signal control scheme.This method can function well even under unfavorable low-saturation conditions.On the basis of the previous work,several indicators are constructed based on the license plate recognition data to evaluate the intersection performance at lane and signal-cycle levels.With these indicators,a method to diagnose the causes of intersection inefficiency is put forward thanks to dynamic time regulation and the grey clustering method.Furthermore,the paper explains the long-time delay pattern of vehicles caused by traffic control schemes with the help of shockwave theory.Compared with existing achievements,the proposed methods focus on the comprehensive impact of traffic characteristics on intersection efficiency with good interpretability.This research draws on and improves advanced models from other disciplines,and innovatively introduces Explainable AI and fuzzy theory from the CI field to describe urban road networks concerning the network’s intrinsic features,such as complexity,dynamics,and fuzziness.Based on these theories,a set of methods is proposed to recognize critical elements in road networks online and reveal their characteristics contributing to their unsatisfactory efficiency.Making the best of data-driven and mechanism models,these methods highlight interpretability and operability,and their feasibilities are illustrated by data experiments.The results are revelatory for managers to deal with the inefficiency of the urban road network.