Traffic control along an urban rail corridor with closely spaced stations can be considered as a sequence of decision-making stages. When a train on an urban rail corridor connecting two terminal points with a large number of intermediate stations is considered, it can follow various regimes of motion and stopping, which identify individual driving scenarios. Their execution may result in different values of attributes describing driving scenarios, namely travel time, energy consumption, and passenger comfort and others.

The objective of this paper is to develop decision rules for driving scenarios along an urban rail corridor which can optimize travel time, energy consumption and passenger comfort, using the concept of machine learning. Machine learning is a science dealing with the development and implementation of computational models of learning and discovery processes. In this paper, the concept of knowledge acquisition through inductive learning as an NHS approach is explored to establish decision rules.

A computer simulation model REGIME was developed for the estimation of values of evaluation criteria, which included travel time, energy consumption and passenger comfort levels. REGIME was used to estimate these values for a hypothetical rail corridor for various driving scenarios. Next, a commercial learning system ROUGH was used in conjunction with the examples created through REGIME to develop decision rules. The learning algorithm used in ROUGH is based on the theory of rough sets proposed by Pawlak.

The study demonstrates the feasibility of machine learning in automated knowledge acquisition to develop decision rules for complex engineering problems such as urban rail corridor control. The technique of machine learning appears to complement the emerging IVHS area. Further research is needed to verify the rules developed before these rules can be applied.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £60 +P&P)