Research Of An Empirical Markov Chain Evolution Algorithm For Designing Driving Cycles Of Floating Car Low-frequency Data

In the traditional data acquisition process for designing vehicle driving cycles,whether the on-board measurement method or the chase-car method,is faced with the problem of a large sampling amount and a high collection cost.With the development of localization technology and mobile devices,the popularization of intelligent terminals,many floating cars(mainly taxies and buses et al.)in cities are equipped with General Packet Radio Service(GPRS)system,which can provide a large number of driving data,save time and cost of data acquisition,and bring new opportunities for designing urban driving cycles.However,the data collection frequency from the floating car is low and unfixed,so it is impossible to design driving cycles by directly employing traditional micro-trips combination optimization method and Markov chain design method(MC).How to use the floating car driving data design driving cycles makes the researchers faced with new challenges.This thesis is devoted to solving the problem of designing driving cycles for floating car low-frequency data.The research is mainly reflected in the following aspects:1.The empirical state transition matrix information of vehicle driving cycles was constructed.In order to explore the transient information expression of the original cycles for low-frequency data,based on characteristic parameters extracted from typical diving cycles,the principal component analysis and K-means clustering were adopted to obtain three kinds of driving cycles which represent the vehicle driving characteristics in different traffic conditions.There are obvious differences in states and transition probability for different kinds of TPMs and the TPMs of the same kind have strong similar characteristics.Based on the similarity of driving cycles of the same kind,the empirical state transition matrix information was constructed,which provided a feasible region to express the state and transition relationship of the original transient for the low-frequency sampling data.2.A Markov chain evolution algorithm for designing driving cycles has been proposed.In order to solve the problem of low design efficiency and low design precision of the MC,further to provides an evolutionary mechanism for designing driving cycle of low-frequency data,a Markov chain evolution method(MCE)was proposed in this work.By combining sampling with evolution,it reconstructed a new genetic algorithm(GA)by using the MC essential characteristics of driving cycles.The mutation and crossover operators which satisfied Markov property were designed.Compared with the two-parameter and three-parameter driving cycles from MC method,the design efficiency of the MCE method increases 53.69 times and 44.50 times,respectively and the better cycles are obtained.3.An empirical Markov chain evolution method for designing cycles of low-frequency data has been proposed.Aiming at the fact that the current design methods is not applicable when faced with sampling frequency of less than 1 Hz,an emiprical Markov Chain Evolution method(EMCE)was proposed based on the empirical state transition matrix information and the MCE method.The emiprical state transition matrix information was constructed to provide a feasible field of the original information for low-frequency cycles.The evaluation indices as objectives were determined by theoretical and statistical analysis for designing cycles of low-frequency data.Finally,the MCE was employed to evolve the desired cycles of the fixed intervals sampling.It was concluded that the EMCE method can design the representative driving cycles for different road types(urban,suburb and highway).The results show that the reconstructed cycles,the original transient cycles and the representative cycles from the typical method have a high consistency on the VA distribution,power spectrum density(PSD)and fuel consumption per 100 km.4.The EMCE was applied to design diving cycles of the floating car in urban.The EMCE method was applied to floating car data(FCD)in Changchun city.Aiming at the change from fixed interval sampling to unfixed interval sampling,the interpolation databases of low-frequency data with fixed and non-fixed interval sampling were respectively obtained by the resampling technique.The related index of VA distribution was used to evaluate the similarity and consistency of two distributions,which was used for analyzing the influence of different interval sampling distributions on the EMCE.The consistency trend of the index corresponding to the mean interval sampling and the fixed interval sampling were taken as the criteria.The maximum interval of 29 s used for designing FCD cycles is determined and the representative driving cycle of FCD collected from Changchun city is obtained.