Timetable multi-objective optimization by improved genetic algorithm
based on stochastic passenger flow
Xubin Sun
1
, Hong Lu
1
, Hairong Dong
2,†
, Jing Xun
2
Abstract— This paper mainly focuses on the method of
timetable multi-objective optimization considering the random-
ness of passengers flow. The passenger flow model is established
based on the normal distribution of the passenger flow, which
is verified through the testing hypotheses for the number of
passengers getting on and off the train. In assessment of
energy consumption, the train operation between the stations
is assumed as the optimal strategy, which is described as four
stages according to Pontryagin maximum principle: maximum
acceleration, speed holding, coasting and maximum braking.
Then an improved GA (genetic algorithm) is designed to obtain
the optimal solution. Finally, this paper test and verify the
effectiveness of the optimal model through the simulation,
taking the running time, departure interval, dwell time as
decision variables, based on the experimental data from the
Beijing subway with the conventional GA and the improved
GA respectively.
I. INTRODUCTION
Urban rail transit has been a significant part for the
development of the country, due to its advantages of high-
capacity, high-speed, high-efficiency from the viewpoint of
development practice of the international metropolis. At
present, there are two common approaches to the multi-
objective optimization. On the one hand, combining the in-
dividual objective functions into a single composite function
is adopted by most people. On the other hand, selecting an
entire Pareto optimal solution set or a representative subset
is also used by some researchers [1]. A large numbers of
researches about the optimization of timetable are aimed at
reducing the operation cost and the waiting time for the
passengers by assuming that the demand of passenger is
constant. However, the robustness of the timetable based on
the stochastic passenger flow should be considered. Cury [2]
proposed a methodology for generation of optimal schedules,
in which a traffic model is represented by the train and
passenger movements. Lasdon [3] and Tamura [4] proposed
a method that use an iterative hierarchical multilevel decom-
position to solve the problem of a string of trains. But this
method is also restricted to small strings of trains, which can
not solve the problem of large dimension. In 2000, Assis and
*This work is supported by projects (61304196), National High Technol-
ogy Research and Development Program of China (2012AA041701).
1
Xubin Sun and Hong Lu are with the School of Electronic and Infor-
mation Engineering, Beijing Jiaotong University, Beijing 100044, China.
2
Hairong Dong and Jing Xun are with State Key Laboratory of Rail
Traffic Control and Safety, Beijing Jiaotong University, Beijing 100044,
China.
†
Corresponding author. Email: hrdong@bjtu.edu.cn
Miani [5] took consideration of all the operational and safety
constraints into the model and use the linear programming
approach to optimize the schedule, assuming that the demand
of passengers is constant. However, the passenger flow in the
management of urban rail transit has a link with the dwell
time from the research of Lam and Cheung [6]. In 2014,
Liu [7] proposed a novel multi-objective train adjustment
model based on the dwelling time model and improved the
PSO (particle swarm optimization) to solve the problem.
However, the exchange behavior of passenger is stochastic,
and is hard to predict, which has a significant influence on
the performance of train. The distribution of passenger has
such significant influence on the train operation and schedule
that the effect of stochastic passenger flow that should not be
neglected. Zuo [8] divided the stations into muster and non-
muster ones, and designed the hence constraints objective
programming models and algorithms based on the stochastic
passenger flow, which revealed the rules of passenger train
layout and flow distribution optimization and formed into rel-
atively complete system. On the other hand, more and more
researchers begin to pay attention to the use of regenerate
braking energy. Sun [9] proposed a multi-train cooperation
method to adjust the train’s profile selected to absorb the
other train’s braking regenerative energy.
As for the method of optimization, more and more re-
searchers tend to intelligent algorithm for the advantages
of flexibility, which have obtained much achievement, such
as the GA, ACO (ant colony algorithm), PSO and so on.
Chang and Kwan [10] used three evaluation of evolution-
ary algorithms, including GA, PSO, and DE (differential
evolution) to solve the scheduling problem. Yao and Yuan
[11] proposed a novel immune genetic algorithm based
on artificial immune algorithm and genetic algorithm, and
apply it to solve the scheduling optimization model for two-
way marshalling train. Being a population-based approach,
GA are well suited to solve multi-objective optimization
problems [1]. Unlike the hill-climbing and gradient descent
search algorithms, the prominent characteristic of GA is
the global search capability [12]. Zhao [13] proposed an
improved genetic algorithms that can avoid the flaw of pre-
cocious and improve global search capability of algorithm.
Chen [14] proposed a new multi-objectives GA without
increasing basic Pareto method’s computation complexity.
As the problem of timetable optimization includes a large
number of variables, the improved GA will be used in this
paper.
2016 IEEE 19th International Conference on Intelligent Transportation Systems (ITSC)
Windsor Oceanico Hotel, Rio de Janeiro, Brazil, November 1-4, 2016
978-1-5090-1889-5/16/$31.00 ©2016 IEEE 434