Study On Dynamic Traffic Guidance Technology Based On Massive Trajectory Data

In the big data era,traffic data has been subjected explosive growth.The acquisition of traffic data is no longer limited to the traditional fixed detector,smart phones,intelligent terminals and other mobile intelligent devices are becoming new traffic data sources as an important part of traffic big data.With the change of trajectory data size,the research contents and methods of trajectory data should be changed.Mass,real-time,high-frequency trajectory data provides an abundant data resources for the dynamic traffic guidance.This study focuses on the technologies of the dynamic traffic guidance based on massive trajectory data.The overall goal of this paper is to use large-scale real-time trajectory data to establish of new data processing algorithm and implementation scheme,to estimate real-time traffic condition accurately and predict traffic flow dynamically,and then to dynamically guide the vehicles accurately.The main achievements of this study include the following aspects:(1)Real-time queue length estimation for signalized intersectionsTo estimate time-dependent queue length problem at signalized intersection is a challenging task for a long time,which will be more difficult if fixed-location sensors are unavailable at these intersections.In this paper,we studied the key technologies on how to use GPS data to estimate real time queue lengths at signalized intersections.We proposed a real time queue length estimation model based on traffic shockwave theory and first put forward the concept of "discharging delay".Utilizing the real time estimation model and the "discharging delays" data of a few queuing vehicles,we can compute real-time queue length estimations cycle-by-cycle.Based on the results of real-time estimation,this paper proposed a guidance-oriented method for calculating the average delay time.In the end we tested the model and algorithm in a real intersection of Guangzhou city.The testing results show that the proposed model and algorithm can estimate queue lengths with satisfactory accuracy.(2)Short-term traffic flow prediction for traffic guidanceShort-term traffic flow prediction is one of the core technologies to realize dynamic traffic guidance.The real-time and accuracy should be taken into account when predicting short-term traffic flow for guidance.Exponential smoothing model has the characteristics of high computational efficiency and is easy to implement,which is suitable for real-time calculation of large-scale data in dynamic induction.However,the traditional exponential smoothing model has the defects of fixed parameters and rigid prediction process.To optimize exponential smoothing model,this paper proposed the Variable Time Domain DynamicModel(VTDDM)for short-term traffic flow prediction.The constrained cyclic coordinate method is used as an optimization algorithm to realize the automatic update of the smoothing coefficient and the data search range according to the change of the measured data in the forecasting process,so that the model can track the trend of the data change so as to realize prediction accuracy.In tests that used field data,the proposed method outperformed the traditional exponential smoothing models on accuracy and reliability.This paper analyzed the relationship between the smoothing coefficient and the data search range,and the results show that they are significantly negatively correlated.(3)Vehicle speed guidance at signalized intersectionsIn order to reduce the intersection delay of the vehicle,this paper studies the vehicle speed guidance at the signalized intersection.Based on traffic shockwave theory,this paper proposed the ideal state of the vehicle arriving at the intersection.On this basis,using the GPS data of the vehicle and the intersection signal information,this paper presented the corresponding vehicle speed guidance algorithm for the different signal states of the vehicle arriving at the intersection.The test results of the multi-agent simulation platform show that the algorithm proposed in this paper can effectively reduce the stopped delay of the vehicle at the intersection and improve the running efficiency of the intersection.(4)Dynamic route guidance at the regional levelThis paper studied the dynamic route guidance at the regional level.Firstly,we proposed the mathematic model for dynamic traffic network and then based on this model we designed the algorithm for dynamic shortest path problem considering turn penalties.Further,utilizing the Big Data technique of HaLoop framework,we designed a distributed and parallel processing model for solving the dynamic shortest path problem.Finally,the algorithm was implemented on the Intelligent Traffic Management and Control Platform Based on Continual Flow,and their computational performance was experimentally evaluated and tested.These results show the guidance algorithm and big data-based parallel computing model can realize dynamic traffic guidance in large scale dynamic network while meeting the real-time requirement.In short,this paper makes a more comprehensive and in-depth exploration of the dynamic traffic guidance problem based on real-time massive trajectory data.The research results obtained in this paper are of great reference and significance for solving the problem of urban traffic congestion.