Arterial Signal Optimization Based On Traffic Arrival Pattern

Intelligent traffic signal optimization is able to promote efficiency in urban traffic control and management,prevent and reduce unnecessary traffic congestion.One of the distinguished features of intelligent traffic signal optimization is to realize data-driven applications.Due to the limitations in the detection range and application costs,traditional point detectors are still problematic to detect traffic arrival and over-saturated traffic demand,as well as guarantee a high information coverage.This makes most of existing signal optimization methods difficulty to apply in field.With the emerging of mobile-internet environment,vehicle trajectory data is becoming a promising traffic flow data source.With the advantages in providing traffic spatiotemporal information and higher information coverage on urban roads,the vehicle trajectory data provides opportunities to overcome the drawbacks of point detection,and develop new approaches for signal optimization.In this context,this study aims to leverage the currently available vehicle trajectory data to obtain traffic arrival information,and develop new signal optimization approaches under both under-and over-saturated conditions.The research work of this study involves four parts: characteristic analysis of vehicle trajectory data,arterial traffic arrival pattern estimation,and arterial signal optimization methods under both under-and over-saturated conditions.The details of the research work are summarized as below.For data characteristic analysis,two types of vehicle trajectory data are primarily considered as they are commonly available on arterials in the cities of China under existing mobile internet environment.One is the license plate recognition(LPR)data,and another is the vehicle location tracing(VLT)data.The purpose of this research is to reveal the data characteristic which is required to develop and validate the traffic arrival pattern estimation method.For LPR data,with a matching process of LPR data collected at two adjacent intersections,the distributions of matching rates over different pairs of adjacent intersections was analyzed.In addition,consider the time drift issue of LPR data relative to the real-time signal phases,a drift-mending method was also developed.For VLT data,as the source platforms of VLT data are not fully shared yet,take the taxi VTL data for instance,the distribution of penetration rates of taxies was analyzed over different time periods of day and roads.In the research of traffic arrival pattern estimation,the traffic arrival pattern is recognized as the change patterns of arrival rates of merge movements over signal cycles,and a lane-based arrival rate estimation method is proposed using vehicle trajectory data.LPR data is used as the main data source,while VLT as the supplementary data when LPR data is not available for certain movements.The method is designed to provide traffic arrival information for the signal timing optimization in the next stages.First,a probability-based model was proposed.According to the characteristic of interrupted flow on urban roads,Piecewise Poisson Process was adopted as the assumption of arrival rates over a signal cycle.The likelihood function was explicitly derived according to the number of vehicle arrivals observed between two matched vehicles from LPR data.Second,as the likelihood function is indifferentiable,a MCMC-based approximate inference technique was used to produce the posterior samples of unknown model parameters,which was further used to calculate the mean and uncertainty estimations of arrival rates.Third,the arrival rate estimation method was tested and evaluated through both traffic simulation and field experiments.The results show that,the proposed method well describes arrive rates of merge movements with piecewise pattern,while it is also able to accommodate to other patterns.The proposed method provides accurate estimations of number of vehicle arrivals during an entire cycle or for a particular movement in both under-and over-saturated conditions.Significant improvements in arrival rate estimation was found when VLT data is incorporated to provide vehicle information for movements not monitored by LPR sensors.In the research of signal optimization for under-saturated arterials,an arterial offset optimization method was proposed through vehicle trajectory reconstruction,which takes advantage of the vehicle arrival information provided by arrival rate estimation.The proposed method is designed to address the problem that most existing offset optimization methods failed to account for actual traffic arrivals.First,vehicle arrival times were extracted from the detailed information implied in arrival rates.Second,an enhanced Newell carfollowing model was proposed to precisely reconstruct the vehicle trajectories.The original Newell model was enhanced to output high-resolution trajectories and properly describe the signal phase state.Based on the reconstructed trajectories,a lane-based delay analysis model is formulated.Third,consider the trade-off between the two directions of arterials,a link-based offset optimization model was designed and solved by a parallel genetic algorithm(GA).To address the problem of link-based offset optimization which may introduce bias in optimization results,an iterative optimization strategy was designed for offline application.Fourth,a real arterial was selected as the test sites,and a simulation experiment was conducted to test and evaluate the proposed model.In the simulation experiment,incomplete data collection environment,historical traffic flow fluctuation and other conditions were comprehensively considered.The simulation results show that,the enhanced Newell model provides accurate and realistic vehicle trajectories.Compared to traditional GA algorithm,the parallel GA is 15 times faster in solving the offset optimization problem.Considering the scenario of iteratively optimizing offsets every week,after 2 iterations(e.g.2 weeks),the optimal offsets would become relatively stable,and the bias caused by link-based strategy can be largely eliminated.Compared to the current signal plan,the optimal offset settings after 2 iterations could reduce the weighted delay of through movements by 13.7%~18.6%.In the research of signal optimization for over-saturated arterials,an arterial metering control method was proposed,taking advantages of the traffic demand information and its distribution over different merge movements provided by arrival rate estimation.The proposed method aims to overcome the problematic assumption of existing methods that traffic demand and turn ratio information is always available,ignoring the difficulties to collect accurate traffic demand and turn ratio information with exiting sensors.First,with the traffic demand information,as well as the saturated headways obtained from LPR data analysis,the degree of saturation can be readily calculated to evaluate the intensity of congestion and determine the critical intersections or lanes.Second,with the implied lane-based OD relationship from traffic demand distribution among different merge movements,a metering analysis model was developed.The model is able to analyze the metering effects without knowing the turn ratios.Third,consider the queue length constraints during a metering control period,a metering green optimization model was designed to minimize the deviation between a pre-setting goal and actual state of the critical intersections(e.g.degree of saturation).Fourth,a simulation experiment was designed to test the proposed method by considering the incomplete data collection environment,entire evolution process of traffic congestion,and etc.The simulation results show that,as the increasing in the duration time of metering control,it would become more difficulty to archive effective metering control.The proposed method can successfully reduce the congestion at critical intersections/lanes without causing spillbacks at metering intersections,as well as maintaining a relatively high discharging flow rates.Compared to current signal plans,the duration of spillback at the critical intersection can be reduced by 55%,and the vehicle delay of the conflicted left turning movements can be also significantly reduced.