Real-time Queue Length Estimation Based On Kalman Filter In Mixed Traffic Environment

Real-time accurate queue length is part of the key factors in measuring and improving intersection control.Traditional detection of queue length relies on very limited,isolated and logging data provided by fixed detectors.The development of the Internet of vehicles technology endows the traffic flow with real-time,dynamic and continuous characteristics.For this reason,based on the low-penetration networked vehicle data and information control data,this paper combines the traffic wave theory with the Kalman filter theory to realize the second-level estimation of the queuing length of the signal intersection under the conditions of different road saturation and networked vehicle penetration rate..The specific content is divided into three parts: estimation of maximum queuing length at signalized intersection based on traffic wave theory;second-level queuing length estimation at intersection based on Kalman filter theory and traffic wave theory;verification of queuing length estimation model based on NGSIM data.(1)A model for estimating the periodic maximum queue length at signalized intersections based on the traffic wave theory is constructed.First,based on the straightforward motion characteristics of connected vehicles at intersections,the definition and extraction algorithm model of feature points is established.Then,by reconstructing the vehicle motion process,a model for estimating the maximum queuing length at an intersection based on the traffic wave theory is constructed under usual conditions and under supersaturated conditions.Finally,the validity of the model is checked by constructing a single-link simulation environment through VISSIM.The results show that the established model can effectively estimate the cyclic queue length in a single-link environment.(2)A second-level queuing length estimation model at intersections based on Kalman filter theory and traffic wave theory is constructed.First,the state transition equation is constructed with the input variables of the arrival rate and departure rate of the entrance at the current moment estimated by the traffic wave theory,and the observation equation is constructed with the current number and penetration rate of queuing connected vehicles.Then,a process algorithm and model performance evaluation index for estimating queue length based on the Kalman filter method are proposed.Finally,by calling VISSIM COM interface,MATLAB is used to solve the model.The results show that the estimation errors under different conditions are small,and the fitting degree is fine with the actual queue length.Compared with the baseline model,the model has a better estimation performance under high permeability and saturation conditions.(3)The queuing length estimation model is verified by using NGSIM data.First,the NGSIM Peachtree Street data was selected as the research object,and the data was preliminaries screened and reprocessed.Then,discrete wavelet transform denoting method is used to denoise and reconstruct the data such as vehicle position,speed and acceleration.Finally,the reconstructed data is substituted into the established model,and the second-level queuing length estimation within nine signal periods is effectively realized.Through the analysis of the error of the results,the scientifically and effectiveness of the method for the application of the example data are proved.The above research results are an exploration of a second-level estimation method for intersection queue length in a mixed traffic environment,which can effectively assist traffic engineers to better evaluate traffic signal control systems and lay a theoretical foundation for future urban traffic situational awareness.