Construction of urban driving cycle of lightduty vehiclebased on LLEKM and Markov chain(PDF)
长安大学学报（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]
- Issue:
- 2021年5期
- Page:
- 118-126
- Research Field:
- 交通工程
- Publishing date:
Info
- Title:
- Construction of urban driving cycle of lightduty vehiclebased on LLEKM and Markov chain
- Author(s):
- ZHANG Huiling1; 2; KONG Dexue3; YU Tao2; AO Guchang1; 2; SHAO Yiming
- (1. Chongqing Key Lab of Traffic System & Safety in Mountain Cities, Chongqing Jiaotong University,Chongqing 400074, China; 2. College of Traffic & Transportation, Chongqing Jiaotong University,Chongqing 400074, China; 3. Intelligent Transportation Systems Research Center,Wuhan University of Technology, Wuhan 430063, Hubei, China)
- Keywords:
- traffic engineering; driving cycle; locally linear embedding; Kmeans clustering; Markov chain
- PACS:
- -
- DOI:
- -
- Abstract:
- In order to objectively describe the operating characteristics of vehicles on the road, a construction method of driving cycle based on locally linear embedding, Kmeans clustering (LLEKM) and Markov chain was proposed to address the shortcomings of the traditional urban road standard driving cycle construction method. The application accuracy of the method was verified. Firstly, combining with the road traffic conditions, the measured sample data was denoised and smoothed. According to the kinematics sequence definition, it was divided into 1 754 kinematics sequences and 16 kinematics parameters were selected to construct a characteristic parameter matrix. Secondly, the locally linear embedding algorithm, which was used for dimensionality reduction and adopted to solve the problem of linear structure loss caused by data compression and reconstruction. Thirdly, through the Kmeans cluster analysis, the variability of each driving state time ratio in different types of data set was calculate, and the driving characteristics of the corresponding types of urban road vehicles were obtained. Then within each category, the state transition probability matrix was confirmed according to the statistics of the adjacent driving states, and the kinematic sequence with minimum error of characteristic parameter was selected by using Markov chain, and the typical driving cycle was synthesized according to the time scale. Finally, the driving data of the same lightduty vehicle in Xiamen city within 3 months was taken as an example, empirical analysis was carried out to verify the accuracy of the proposed method. The results show that compared with sample data and driving cycle constructed by traditional methods, the driving cycle constructed by using the proposed method has a better similarity with sample data in the aspects of main characteristic parameters, and the average relative error rate is reduced from 9.69% to 4.47%. The velocityacceleration joint probability distribution error of the proposed method is obviously reduced, and the driving cycle construction accuracy is improved. Comparing and analyzing the standard driving cycle constructed by current methods, this method can better reflect the actual operating conditions of lightduty vehicles in Xiamen. 5 tabs, 8 figs, 26 refs.
Last Update: 2021-09-30