TR2015-063

A Framework for Real-time Near-optimal Train Run-curve Computation with Dynamic Travel Time and Speed Limits


    •  Xu, J.; Nikovski, D.N., "A Framework for Real-Time Near-Optimal Train Run-Curve Computation with Dynamic Travel Time and Speed Limits", American Control Conference (ACC), DOI: 10.1109/ACC.2015.7170790, ISBN: 978-1-4799-8685-9, July 2015, pp. 533-540.
      BibTeX Download PDF
      • @inproceedings{Xu2015jul,
      • author = {Xu, J. and Nikovski, D.N.},
      • title = {A Framework for Real-Time Near-Optimal Train Run-Curve Computation with Dynamic Travel Time and Speed Limits},
      • booktitle = {American Control Conference (ACC)},
      • year = 2015,
      • pages = {533--540},
      • month = jul,
      • publisher = {IEEE},
      • doi = {10.1109/ACC.2015.7170790},
      • isbn = {978-1-4799-8685-9},
      • url = {http://www.merl.com/publications/TR2015-063}
      • }
  • MERL Contact:
  • Research Area:

    Data Analytics


This paper studies the problem to generate the most energy efficient run-curves subject to given travel time requirements. The target is to provide a train with the ability to quickly adjust its run curve according to different travel time requirements and speed limits along the track before departing a terminal. Using a train model considering train length, varying track gradient and speed limit profile, the optimal run-curve problem is formulated into a bi-criteria optimization problem that minimizes weighted energy consumption and weighted travel time. By selecting appropriate weight values, the optimization problem would generate a run-curve with near-optimal energy consumption. We propose a two stage procedure framework, which includes an off-line stage and a real-time stage. A series of geometric relation between weight in the objective function and travel time are derived. The actual run-curves are generated in the real-time stage using approximate dynamic programming. For the first time, a framework provides trains with the ability to fast response to dynamic travel time requirement using complex physical models.