Please wait a minute...
Tsinghua Science and Technology  2019, Vol. 24 Issue (2): 226-237    doi: 10.26599/TST.2018.9010114
A Hierarchical Ensemble Learning Framework for Energy-Efficient Automatic Train Driving
Guohua Xi, Xibin Zhao, Yan Liu, Jin Huang*, Yangdong Deng
∙ Guohua Xi is with the CRRC Corporation Limited, Beijing 100078, China.
∙ Xibin Zhao, Yan Liu, Jin Huang, and Yangdong Deng are with the School of Software and Key Laboratory for Information System Security, Ministry of Education (KLISS)/Beijing National Research Center for Information Science and Technology (BNRist), Tsinghua University, Beijing 100084, China.
Download: PDF (4978 KB)      HTML
Export: BibTeX | EndNote (RIS)      


Railway transportation plays an important role in modern society. As China’s massive railway transportation network continues to grow in total mileage and operation density, the energy consumption of trains becomes a serious concern. For any given route, the geographic characteristics are known a priori, but the parameters (e.g., loading and marshaling) of trains vary from one trip to another. An extensive analysis of the train operation data suggests that the control gear operation of trains is the most important factor that affects the energy consumption. Such an observation determines that the problem of energy-efficient train driving has to be addressed by considering both the geographic information and the trip parameters. However, the problem is difficult to solve due to its high dimension, nonlinearity, complex constraints, and time-varying characteristics. Faced with these difficulties, we propose an energy-efficient train control framework based on a hierarchical ensemble learning approach. Through hierarchical refinement, we learn prediction models of speed and gear. The learned models can be used to derive optimized driving operations under real-time requirements. This study uses random forest and bagging – REPTree as classification algorithm and regression algorithm, respectively. We conduct an extensive study on the potential of bagging, decision trees, random forest, and feature selection to design an effective hierarchical ensemble learning framework. The proposed framework was testified through simulation. The average energy consumption of the proposed method is over 7% lower than that of human drivers.

Key wordsmachine learning      energy efficiency      train driving system      feature selection      ensemble learning     
Received: 20 April 2018      Published: 29 April 2019
Corresponding Authors: Jin Huang   
About author:

Yangdong Deng received the ME and BE degrees from Tsinghua University, Beijing, China, in 1998 and 1995, respectively, and the PhD degree in electrical and computer engineering from Carnegie Mellon University, Pittsburgh, PA, USA, in 2006. He is now an associate professor in the School of Software, Tsinghua University, Beijing, China. His current research interests include parallel electronic design automation algorithms, parallel program optimization, general purpose computing on graphics processing hardware, etc.

Cite this article:

Guohua Xi, Xibin Zhao, Yan Liu, Jin Huang, Yangdong Deng. A Hierarchical Ensemble Learning Framework for Energy-Efficient Automatic Train Driving. Tsinghua Science and Technology, 2019, 24(2): 226-237.

URL:     OR

Fig. 1 Illustration of slope of one railway route.
Fig. 2 A sample of human driving data. Blue and red lines indicate the velocity and driving gears, respectively, for a group of drivers. We draw them together to show that possible patterns exist among the driving operations.
Fig. 3 Proposed solution framework for energy-efficient train driving. REPTree means Reduced Error Pruning Tree.
Fig. 4 First layer for velocity prediction.
Fig. 5 Second layer for gear prediction.
Section typeLableResultant gradient
Steep down grade section-2-3
Gentle down grade section-1-1 to -3
Gentle grade section0-1 to 1
Gentle up grade section11 to 3
Steep up grade section23
Table 1 Route section categories.
Fig. 6 Designed features set for driving records covering train attributes, railway properties, and running information.
M1Prediction of the velocity change combinations of the slope sections
M2Prediction of the proportion of the velocity in velocity change combinations
M3Prediction of the gear change combinations of the velocity section
M4Prediction of the proportion of the gear in gear change combinations
Table 2 Trained prediction models.
Fig. 7 Power characteristics of selected locomotive for brake.
Fig. 8 Power characteristics of selected locomotive for traction.
Fig. 9 Illustration of railway route using in the experiments.
Fig. 10 Photo of hardware-in-loop test platform of the system with: ①-Working condition generator, ②-LKJ2000 train running monitor and record device, ③-displayer of LKJ2000, ④-supplementary communication device from LKJ2000, ⑤-master controller, ⑥-throttle signal converting device, ⑦-power supplier, ⑧-onboard trip optimization&control device, and ⑨-train motion simulation platform.
ModelPrecision (%)Recall (%)F-Measure (%)CCMAERMSERAE (%)RRSE (%)
Table 3 Evaluation results of trained prediction models.
Fig. 11 Illustration of a sample driving trip by proposed approach compared with a human driver record.
No.Load (ton)EC-Driver (kg)TC-Driver (s)EC-Proposed (kg)TC-Proposed (s)ESTDES (%)
Table 4 Comparison of driving performance between drivers and proposed approach.
[1]   Huang J., Deng Y. D., Yang Q. W., and Sun J. G., An energy- efficient train control framework for smart railway transportation, IEEE Trans. Comput., vol. 65, no. 5, pp. 1407-1417, 2016.
[2]   Huang J., Yang F., Deng Y. D., Zhao X. B., and Gu M., Human experience knowledge induction based intelligent train driving, in Proc. IEEE/ACIS 16th Int. Conf. Computer and Information Science, Wuhan, China, 2017, pp. 335-340.
[3]   Ghosh S. and Lee T., Intelligent Transportation Systems: New Principles and Architectures. Boca Raton, FL, USA: CRC Press, 2002.
[4]   Li P., Jia L. M., and Nie A. X., Study on railway intelligent transportation system architecture, in Proc. 2003 Intelligent Transportation Systems IEEE Int. Conf., Shanghai, China, 2003, pp. 1478-1481.
[5]   Han S. H., Byen Y. S., Baek J. H., An T. K., Lee S. G., and Park H. J., An optimal automatic train operation (ATO) control using genetic algorithms (GA), in Proc. IEEE Region 10 Conf. Multimedia Technology for Asia-Pacific Information Infrastructure, Cheju Island, South Korea, 1999, pp. 360-362.
[6]   Li Y. S. and Hou Z. S., Study on energy-saving control for train based on genetic algorithm, (in Chinese), J. Syst. Simul., vol. 19, no. 2, pp. 384-387, 2007.
[7]   Wang Y. H., De Schutter B., van den Boom T. J. J., and Ning B., Optimal trajectory planning for trains—A pseudospectral method and a mixed integer linear programming approach, Transp. Res. Part C: Emerg. Technol., vol. 29, pp. 97-114, 2013.
[8]   Kumar A. K., Daum W., Shaffer G. R., Houpt P. K., Noffsinger J. F., and McKay D. L., Trip optimizer method, system and computer software code for operating a railroad train to minimize wheel and track wear, U.S. Patent 9233696, January 12, 2016.
[9]   Caprara A., Kroon L., and Toth P., Optimization problems in passenger railway systems, in Wiley Encyclopedia of Operations Research and Management Science, Cochran J. J., Cox L. A. Jr., Keskinocak P., Kharoufeh J. P., and Smith J. C., eds. New York, NY, USA: John Wiley & Sons, 2011.
[10]   Miyatake M. and Ko H., Optimization of train speed profile for minimum energy consumption, IEEJ Trans. Electr. Electron. Eng., vol. 5, no. 3, pp. 263-269, 2010.
[11]   Chevrier R., An evolutionary multi-objective approach for speed tuning optimization with energy saving in railway management, in Proc. 13th Int. IEEE Conf. Intelligent Transportation Systems, Funchal, Portugal, 2010, pp. 279-284.
[12]   Su S., Li X., Tang T., and Gao Z. Y., A subway train timetable optimization approach based on energy-efficient operation strategy, IEEE Trans. Intell. Transport. Syst., vol. 14, no. 2, pp. 883-893, 2013.
[13]   Huang J., Chen Y. H., and Cheng A. G., Robust control for fuzzy dynamical systems: Uniform ultimate boundedness and optimality, IEEE Trans. Fuzzy Syst., vol. 20, no. 6, pp. 1022-1031, 2012.
[14]   Huang J., Huang Q. M., Deng Y. D., and Chen Y. H., Toward robust vehicle platooning with bounded spacing error, IEEE Trans. Comput.-Aided Des. Integr. Circ. Syst., vol. 36, no. 4, pp. 562-572, 2017.
[15]   Huang J., Chen Y. H., and Guo K. H., Novel approach to multibody system modeling: Cascading and clustering, J. Aerosp. Eng., vol. 27, no. 2, pp. 279-290, 2014.
[16]   Salmasi F. R., Control strategies for hybrid electric vehicles: Evolution, classification, comparison, and future trends, IEEE Trans. Veh. Technol., vol. 56, no. 5, pp. 2393-2404, 2007.
[17]   Ding Y., Bai Y., Liu F. M., and Mao B. H., Simulation algorithm for energy-efficient train control under moving block system, in 2009 WRI World Congress on Computer Science and Information Engineering, 2009, pp. 498-502.
[18]   Ke B. R., Lin C. L., and Lai C. W., Optimization of train-speed trajectory and control for mass rapid transit systems, Control Eng. Pract., vol. 19, no. 7, pp. 675-687, 2011.
[19]   Wang D. Y., Lin X., and Zhang Y., Fuzzy logic control for a parallel hybrid hydraulic excavator using genetic algorithm, Autom. Construct., vol. 20, no. 5, pp. 581-587, 2011.
[20]   Gao S. G., Dong H. R., Chen Y., Ning B., Chen G. R., and Liang Q. W., Neuro-adaptive and robust automatic train control subject to unknown dynamics, in Proc. 32nd Chinese Control Conf., Xi’an, China, 2013, pp. 8130-8134.
[21]   Mao B. H., Train Performence Calulation and Design. Beijing, China: China Communications Press, 2008.
[22]   Hall A. M., Correlation-based feature selection for machine learning, Ph.D. dissertation, University of Waikato, Hamilton, New Zealand, 1999.
[23]   Yang Q. W., Hu X. Q., Lei X. C., Zhu Y., Wang X. Y., and Ji S. C., Adaptive operation strategy for voltage stability enhancement in active DMFCs, Energy Convers. Manage., vol. 168, pp. 11-20, 2018.
[24]   Yang Q. W., Hu X. Q., Zhu Y., Lei X. C., Yu B., and Ji S. C., Extended criterion for robustness evaluations of energy conversion efficiency in DMFCs, Energy Convers. Manage., vol. 172, pp. 285-295, 2018.
[25]   Hall M. A. and Smith L. A., Practical feature subset selection for machine learning, in Proc. 21st Australasian Computer Science Conf., Berlin, Germany, 1998.
[26]   Liaw A. and Wiener M., Classification and regression by randomForest, R. News, vol. 2, no. 3, pp. 18-22, 2002.
[27]   Kalmegh S., Analysis of WEKA data mining algorithm REPTree, Simple CART and RandomTree for classification of Indian news, Int. J. Innov. Sci. Eng. Technol., vol. 2, no. 2, pp. 438-446, 2015.
[28]   Breiman L., Bagging predictors, Mach. Learn., vol. 24, no. 2, pp. 123-140, 1996.
[29]   Frank E., Hall M., Holmes G., Kirkby R., Pfahringer B., Witten I. H., and Trigg L., Weka, in Data Mining and Knowledge Discovery Handbook, Maimon O. and Rokach L., eds. Boston, MA, USA: Springer, 2005, pp. 1305-1314.
[30]   Kohavi R., A study of cross-validation and bootstrap for accuracy estimation and model selection, in Proceedings of the 14th International Joint Conference on Artificial Intelligence, Montreal, Canada, 1995.
[1] Catalina Aranzazu-Suescun,Mihaela Cardei. Distributed Algorithms for Event Reporting in Mobile-Sink WSNs for Internet of Things[J]. Tsinghua Science and Technology, 2017, 22(4): 413-426.
[2] Yanhua Li,Youhui Zhang,Weiming Zheng. HW/SW Co-optimization for Stencil Computation: Beginning with a Customizable Core[J]. Tsinghua Science and Technology, 2016, 21(5): 570-580.
[3] Yi Yang,Wenguang Chen. Taiga: Performance Optimization of the C4.5 Decision Tree Construction Algorithm[J]. Tsinghua Science and Technology, 2016, 21(4): 415-425.
[4] Jiancheng Zhong,Jianxin Wang,Wei Peng,Zhen Zhang,Min Li. A Feature Selection Method for Prediction Essential Protein[J]. Tsinghua Science and Technology, 2015, 20(5): 491-499.
[5] Boyuan Liu,Wenhui Fan,Tianyuan Xiao. Unsupervised Dynamic Fuzzy Cognitive Map[J]. Tsinghua Science and Technology, 2015, 20(3): 285-292.
[6] Su Xing, Tong Hanghang, Ji Ping. Activity Recognition with Smartphone Sensors[J]. Tsinghua Science and Technology, 2014, 19(3): 235-249.
[7] . Reliable and Energy Efficient Target Coverage for Wireless Sensor Networks[J]. Tsinghua Science and Technology, 2011, 16(5): 464-474.
[8] . Improved Adaptive Compression Arbitration System for Wireless Sensor Networks[J]. Tsinghua Science and Technology, 2010, 15(2): 202-208.
[9] . Soft Sensing of Overflow Particle Size Distributions in Hydrocyclones Using a Combined Method[J]. Tsinghua Science and Technology, 2008, 13(1): 47-53.