Vehicle Trajectory Reconstruction Model Based On Acceleration Distributions For Emission Estimation

Road vehicle emissions have become one of main sources of the air pollution in urban areas.In the effective control of urban air pollutions,an accurate emission estimation of pollutants is the key to the evaluation of the effects of policies on energy reduction and emission mitigation.The estimation of emissions depends on the development of emission models.For the widely used emission models such as Motor Vehicle Emissions Simulator?MOVES?,the Vehicle Specific Power?VSP?and its distributions are the key parameters for the emission estimation.The development of VSP distributions involves a strict data requirement,i.e.it has to be provided with second-by-second vehicle trajectories.At present,the GPS data collection and microscopic traffic simulation are the two main sources of data for developing VSP distributions.However,there are two practical problems when using these two data sources to develop VSP distributions:?1?The data storage and transmission for the collection of large amounts of second-by-second vehicle trajectories are prohibitively expensive.Thus,it is unrealistic to accomplish the real-time transmission of large amounts of second-by-second trajectories in light of the limitation of the present computer technology.On the other hand,there exists abundant sparse vehicle trajectory data?i.e.where the data interval is larger than one second?for the purpose of navigation or traffic state surveillance.However,these sparse data cannot be used directly for the accurate calculation of acceleration and VSP,thus,they cannot be used for accurate emission estimation.?2?Although microscopic traffic simulation models can generate second-by-second vehicle trajectories,these models were originally developed for the estimation of aggregated traffic parameters,such as average speed and time headway,at the expense of the ability to accurately capture second-by-second acceleration characteristics,resulting in a low precision of emission estimations based on simulated trajectories.In this context,the research in this dissertation is intended to develop trajectory reconstruction models for the field observed sparse trajectories as well as simulated trajectories from the microscopic simulation model for urban expressways.The main purpose of the research is to resolve the issue regarding how the field observed sparse trajectories and simulated trajectories be effectively used to accurately estimate emissions.A summary of main findings for the research is provided as follows:?1?The effect of acceleration stochastic characteristics on emission estimations was studied.Two interpolation methods,the linear interpolation and the cubic spline interpolation,which do not consider the acceleration stochastic characteristics were used to interpolate the sparse trajectories.The maximum relative emission errors of the sparse trajectories interpolated by the linear interpolation and the cubic spline interpolation were as high as 33.3%and 17.3%respectively.It was concluded that the trajectory interpolations,which does not consider the acceleration stochastic characteristics,will lead to the inaccurate emission estimation.In order to accurately describe the acceleration stochastic characteristics,this thesis proposed the basic method of using statistical acceleration characteristics,i.e.acceleration distribution,to describe acceleration stochastic characteristics.?2?Based on large amounts of observed second-by-second trajectories of light-duty vehicles,a method of describing the stochastic characteristics of the acceleration based on the models of acceleration distribution at different operating states was proposed.Firstly,according to the variation range of vehicle speed,three operating states were defined:steady state,unsteady-accelerated state and unsteady-decelerated state.Secondly,a method for identifying the three operating states was developed based on its definition,and the characteristics of acceleration distributions at the three corresponding operating states were analyzed.Finally,the acceleration distribution models were developed by fitting the probability density functions of acceleration distributions at the three operating states.Precisely,the goodness-of-fit of these acceleration distributions were all above 0.9.?3?A trajectory reconstruction model for the sparse trajectories was developed based on the stochastic characteristics of the acceleration and acceleration distribution models.Firstly,the causes of emission estimation errors based on sparse trajectories were analyzed.Secondly,the sparse trajectories were divided into fragments by every30 seconds,and the operating states of each trajectory fragment were identified.According to the identified operating states,the corresponding acceleration distribution model was selected to generate the random accelerations to reconstruct the trajectories between the sparse points of each trajectory fragment.This method was termed the interpolation based on acceleration distribution?ICD?in this dissertation.Finally,the effectiveness of the ICD method was validated by a comparison with the existing cubic spline interpolation?CSI?.It was shown that,for the sparse trajectories with an interval above five seconds,the relative errors of emission factors for CO₂,CO,NOx and THC by ICD can be reduced by up to 8.5%?26.9%?21.3%?24.4%respectively in a comparison with CSI.For the sparse trajectories with an interval under five seconds,the relative errors of the emission factors by CSI was slightly lower than that of ICD,but the difference between the relative errors of emission factors by ICD and CSI was within 2%.?4?A trajectory reconstruction model for the microscopically simulated trajectories was developed based on the stochastic characteristics of the acceleration and acceleration distribution models.Firstly,an analysis to the causes of the errors of the emission estimation based on simulated trajectories was analyzed.Secondly,the reconstruction mechanism of simulated trajectories was provided by extracting aggregated characteristics of traffic parameters of the microsimulation,based on the stochastic characteristics of the acceleration and acceleration distribution models.Finally,using the simulated trajectories from the microscopic simulation model VISSIM as a case platform,a trajectory reconstruction model was developed for the simulated trajectories.The relative errors of emission factors for CO₂,CO,NO_X and THC after the reconstruction were shown to be reduced by up to 22.2%,59.3%,14.3%and 16.5%respectively.?5?In the case study using a real expressway section,the developed trajectory reconstruction models for both sparse trajectories and simulated trajectories were validated respectively.It was observed that the proposed trajectory reconstruction models for both sparse trajectories and simulated trajectories could effectively reduce the emission estimation errors.Regarding the reconstruction of the sparse trajectories,the average relative errors of CO₂ emissions based on the sparse trajectories interpolated by ICD for any intervals within 30 seconds were all within 5%.Regarding the reconstruction of the simulated trajectories,the relative errors of CO₂ emissions at steady state,unsteady-accelerated state and unsteady-deaccelerated state based on the proposed reconstruction method could be reduced by up to 34.6%,13.7%,and 30.8% respectively.