Professional Documents
Culture Documents
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2020.Doi Number
ABSTRACT Transit-Oriented Development (TOD), which integrates the Urban Rail Transit (URT) and
surrounding land-use planning, is a noteworthy way to achieve sustainable development of URT. Stations
with different function orientations have different development goals. However, there are limited studies
that focus on the TOD planning with distinct station types. This study aims to propose a method to fill this
gap. Commercial and residential development types are two dominating Station-TOD types. In this study, a
multi-objective TOD planning model which integrates urban rail transit station, commercial, residential and
public land-uses is proposed. The objectives focus on land value, ridership of URT, the compactness and
commercial accessibility of land cells. Meanwhile, a two-phase solution algorithm based on genetic
algorithm is designed to search for a set of Pareto solutions of alternative land-use plans. Two URT stations
in Beijing are taken as case studies. One is Sidaoqiao station as a commercial-dominating type case, the
other one is Pingxifu station as a residential-dominating type case. The results show that the optimized plan
is more reasonable than the current plan. Both of the cases demonstrate the effectiveness and superiority of
the proposed model and algorithm.
INDEX TERMS Urban rail transit station, Transit-oriented development, Land-use, Sustainable development.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
environmental carrying capacity. Then, we propose a two- of Wuhan in China, showing that a higher transit proximity
phase solution approach based genetic algorithm (GA) to premium of housing prices is most likely to occur in the
search for an optimal land-use solution. Finally, two URT place with more mixed land-use. Gu et al. [18] studied the
stations in Beijing are taken as cases. One is Sidaoqiao air quality in 37 Chinese cities, found that transit-oriented
station as a commercial-dominating type case, the other one development is linked with better air quality. Furthermore,
is Pingxifu station as a residential-dominating type case. TOD research perspectives include regional-level and
We analyze the final results and discuss the comparison station-level [19], [20]. From the perspective of station-
between the optimal solution and current plan. level, Lyu et al. [21] classified the 268 metro station areas
in Beijing into six distinct types. Except for the station-
level study, Sahu [22] also established a TOD model which
aims to decide the land types and densities along a transport
corridor.
B. MATHEMATICAL MODEL
In addition, the research on the analytical models for TOD
planning is also a concentrated field. TOD land planning
model comes down to a multi-objective optimization
problem [23]. The objectives and constraints are functions
of the decision variables which include the land-use type
and density [24]. TOD built environment variables include
dependent and independent variables. The dependent
variables include metro ridership, metro ridership
dispersion overtime. The independent variables include
residential density, employment density, building density,
FIGURE 1. The concept of TOD [8].
land-use variety [15]. Lin et al. [25] proposed a TOD
The rest of this paper is organized as follows. Section 2 planning linear continuous model considering the limited
reviews the recent literature. Section 3 describes the land-use densities. The objectives of the model include:
proposed problem, the design framework and the multi- maximizing the urban rail transit ridership, improving
objective programming model is established. Furthermore, living-environment quality and maximizing social equity of
a two-phase solution algorithm based on GA is designed to land planning. The constraints of the model include limiting
search for a set of Pareto solutions in Section 4. Section 5 land-use density, land-use variety. Besides, the objective of
describes two case results to demonstrate the effectiveness maximizing the land value around station and the constraint
and superiority of the proposed model and algorithm. of pollution cost control was also considered in a TOD
Finally, the conclusions and future work are presented in model built in [26]. The accessibility around the station was
the last section. taken as an objective in [20]. Genetic Algorithm (GA) is the
primary method to solve multi-objective, nonlinear,
II. LITERATURE REVIEW
complex integer programming model, it can help planners
to get alternative land-use plans [27], [28]. A Grey TOPSIS
A. TOD STRATEGIES AND PRINCIPLES
There have been many studies highlighting TOD strategies method was also applied in [19]. The method of estimating
and principles. TOD highlights medium to high compact, the key parameters in an integrated land-use and
mixed land-use, walkability, and centered on public transit transportation model was proposed in [29].
[8]. Density, Diversity, and Design are three essential In summary, many qualitative researches have been done
dimensions of traditional TOD planning [9]. Then Ewing et on the strategies and principles of TOD planning. However,
al. [10], [11] added accessibility, distance to transit and the research on the TOD mathematical model is not enough,
demand management to upgrade “3Ds” to “6Ds”. TOD especially the models on station-level. Besides, different
integrates urban development and transport, mainly through station types focus on distinct TOD strategies, thus the
increasing the ridership by raising land density, improving objectives and constraints of the station land-use planning
the convenience by diversifying the land-use types, and models are diverse based on its type. However, little prior
increasing the attractiveness by improving traffic conditions work takes the station types into account. TOD planning
[12]-[14]. The latest research in this area focused on the emphasizes on differentiated development, the closer to the
TOD impact on transit ridership, housing price and station center, the higher density and vice versa [30]. In the
environment. The transit ridership is influenced by TOD previous models, gradient density was not considered either.
strategies [15], [16]. Sung found transit ridership is
III. MODEL FORMULATION
positively affected by the business and residential density
and land-use mix index [16]. Li et al. [17] used a case study
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
This paper aims to establish a model integrates transport and Besides, this study assumes that each cell has only one
land-use plan of URT stations for transit-oriented land-use type and the density of cells with same type and
development. The land-use plan needs to decide the quantity, zone are identical.
location and density of each land-use type cell in station area.
Different land-use plans have different TOD effects.
Meanwhile, stations with different TOD types have different
development goals depending on their functional orientation.
Commercial-dominating and residential-dominating station
types are two main TOD station types. Thus, this study aims
to build a model searching for the best land-use plan of these
two types of stations.
A. DESIGN FRAMEWORK
The reasonable walking and bicycle distance from the station
is considered as the radius of TOD scale. This paper takes the
average acceptable walking and cycling distances for 10
minutes, and the study area of TOD stations is a circular area
with a radius of 900m [30]. Then the study area is divided
into 900 square cells of 60m×60m considering the accuracy
of model and feasibility of solution algorithm. The solution
result of the model is very sensitive to the impacts of cell
length. If the cell length is too long, the model will not
accurately describe the land-use status. On the contrary, the
FIGURE 3. Coordinate system of study area.
shorter cell length leads to too many cells, affecting the
solution efficiency. Meanwhile, this paper investigates the
B. NOTATIONS DEFINITIONS
house prices along the Beijing URT Batong line and
Yizhuang line, the relationship between house price and TABLE 1. Variables used in the paper.
station-TOD driving effect can be analyzed (see Fig. 2). Then
the study scope can be divided into three zones considering Variables Definition
the gradient development (see Fig. 3). Cell number of row and column in study area, 1 ≤ i, j
i, j
≤ 30
Zone division in study scope, Z={Z1,Z2,Z3}, Z1 represent
Z internal zone, Z2 represent intermediate zone, Z3
represent external zone
Types of land-use, which is coded in this paper as
follows: commercial and office as 1; residential as 2;
k public space as 3; green space as 4; transport facilities
as 5; pending to develop as 0; out of the study scope as
9
Land-use type index of cell (i, j) for land-use k, which is
X ijk a binary variable that equals 1 if land-use type of cell (i,
j) is k and equals 0 otherwise, k={1,2,3}
rijk Land-use density of type k at cell (i, j), k={1,2,3}
S Area of cell (i, j), 60m×60m=3600m2
Total number of cells that can be developed in the study
FIGURE 2. Relationship between house price and distance to station N
scope
center.
NB Total number of commercial cells in study scope
1) Internal zone: 0-180m, there are more commercial and
N B1 Total number of commercial cells in internal zone
office buildings, public transport facilities, public open space
and fewer residential buildings. NB2 Total number of commercial cells in intermediate zone
2) Intermediate zone: 180m-480m, “walking comfort NP Total number of public space cells in study scope
zone”. The residential house price in this zone is significantly NR Total number of residential cells in study scope
higher than other zones. This zone belongs to a significant CB Compactness of commercial cells
value-added zone. CR Compactness of residential cells
3) External zone: 480-900m, the residential house price in Lij The distance between cell (i, j) and station center
this zone has dropped significantly, and it becomes
inconvenient for surrounding residents due to the increased
travel time.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
Value
Symbol Definition Notes
Commercial type station Residential type station
32,040,000 43,884,000 Z1
M ij1 Land price of commercial use at cell (i, j) 28,803,600 39,423,600 Z2
21,384,000 29,268,000 Z3
31,536,000 43,236,000 Z1
M ij2 Land price of residential use at cell (i, j) 31,536,000 43,236,000 Z2
28,908,000 39,632,400 Z3
M ij3 Land price of public place at cell (i, j) 7,200,000 9,720,000
817 person/10,000 m2 k=1
δ ak Trip attraction rate of cell for land-use type k 171 person/10,000 m2 k=2
250 person/10,000 m2 k=3
621 person/10,000 m2 k=1
δ gk Trip generation rate of cell for land-use type k 400 person/10,000 m2 k=2
200 person/10,000 m2 k=3
0.14 k=1
τk Peak hour coefficient of land-use type k 0.12 k=2
0.06 k=3
ϑ External travel coefficient of residents in study scope 0.75 0.93
ω Urban rail transit mode share rate 0.472
Z1 / Z2 / Z3 Z1 / Z2 / Z3
Upper bound of density at cell (i, j), which is related 3.5/3.3/2.9 3.3/3.1/2.9 k=1
rijk (u )
to type k and located zone 2.8/2.6/2.2 2.6/2.4/2.2 k=2
2.0/1.8/1.4 1.8/1.6/1.4 k=3
2.1 k=1
Lower bound of density at cell (i, j), which is only
rijk (l ) 1.6 k=2
related to type k
0.8 k=3
εu Upper bound of per capita public space 4.213m2/person
εl Lower bound of per capita public space 2.218 m2/ person 1.8m2/ person
60m2/ person k=1
ϕk Per capita floorage of land-use k
33.08m2/ person k=2
χ Employment coefficient of resident population 0.462
JRRu Upper bound of job-housing balance 1.2
JRRl Lower bound of job-housing balance 0.8
65.84% 23.36% k=1
α ku Upper bound of proportion for cells with type k 69.31% 95.29% k=2
19.80% 18.24% k=3
29.95% 0 k=1
α kl Lower bound of proportion for cells with type k 22.77% 68.85% k=2
1.73% 0 k=3
nr Number of trains operating in peak hour 10 (Line S1) 20 (Line No.8)
cap Train capacity 900 person (Line S1) 1440 person (Line No.8)
(Data source: Beijing Municipal Commission of Planning and Natural Resources. [31]-[33])
the station, the accessibility of the cell is better and the price
C. OBJECTIVES is higher. For greater land value-added with the limited land
The core strategy of TOD is to integrate transportation and resources in the study scope of the station, it should make a
urban development to promote economic development and reasonable planning of land layout. For example, make the
urban progress. Thus, the objectives of the model mainly commercial lands near the station.
include: maximize the land value, ridership, land-use Because the total exploitable cell number is unique in the
compactness and commercial-use land accessibility through study scope of each station, it is not applicable to compare
reasonable land-use design. Commercial-dominating and with the total land value between stations. In this paper, the
residential-dominating type stations focus on different average land price of all cells is proposed, we calculate the
objectives, the models are established by the combination of total value of all cells within the study scope and divide by
these objectives. the total number of cells. It can be formulated as follows:
1) MAXIMIZATION OF LAND VALUE
TOD can increase land price in station area. Land with ∑ ∑
( i , j )∈Z k =
1,2,3
X ijk M ijk
different land-use types will have a different appreciation in max Z1 = (1)
NB + NR + NP
diverse locations around the station. The distance between
the cell and station influences the land price. The closer to 2) MAXIMIZATION OF URT RIDERSHIP
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
The sustainable development of URT needs huge to increase the accessibility between the working place and
passenger flow [25]. What the passenger flow referred to in the rail transit station for attracting more passenger flow [20],
this objective is the total URT ridership generated and [25]. The shorter distance between the commercial cell and
attracted within the station study scope in peak hour [34]. station, the better accessibility. Minimum of the average
First, the total generated and attracted transport ridership is distance between commercial cells and station center can be
calculated according to the land-use type, density, area of the formulated as follows:
cell. Then, URT ridership is assigned according to the peak
hour coefficient, external traveling rate and URT model share ∑L ij ⋅ X ij1
, ∀(i, j ) ∈ {Z1 , Z 2 }
(i , j )
rate. Maximum URT ridership can be formulated as follows: min
= Z5 (5)
N B1 + N B 2
max Z 2 =ϑ ⋅ ω ∑ ∑ τ k S ⋅ X ijk ⋅ rijk (δ ak + δ gk ) (2)
(i , j )∈Z k =
1,2,3
D. CONSTRAINTS
The nature of the multi-objective optimization problem is to
3) MAXIMIZATION OF LAND-USE COMPACTNESS find the optimal solution in a limited range. The constraints
A group of residential lands usually forms a residential of the TOD model are related to the independent variables of
community for improving land-use efficiency and living land-use design which conform to the basic principles of
convenience [20]. The third objective aims to maximize the urban planning. For example, the upper and lower limits of
residential land-use compactness in the study scope. density should be set. The density in station area should
However, commercial lands need special consideration. In ensure gradient development as well. Besides, in a TOD
the internal and intermediate zone, a compact design of community, the quantity of each land-use type cells needs to
commercial lands can improve land value and land be controlled in a reasonable range to guarantee the job-
integration efficiency. In the external zone, the commercial housing balance and enough per-capita public land area.
lands should dispersive to provide better service for the
1) ENSURING APPROPRIATE DEVELOPMENT DENSITY
surrounding residents. The fourth objective aims to maximize
A low-level of development density will be unable to
the commercial land-use compactness in the internal and
ensure higher land-use efficiency and land value. A high-
intermediate zone.
level of development density will attract more ridership.
However, it will bring a series of problems, such as
worsening of working and living environment and causing
the traffic congestion. The density of land development
should meet the environmental capacity and other constraints.
Thus, the land-use density should have its lower and upper
bounds according to the controls on urban planning.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
A good way to reduce tidal passenger flow is to provide optimization problems. It may be impossible to achieve the
adequate jobs for residents in the station area. The best result that satisfies all constraints and makes all
commercial-dominating type stations focus on the balance objectives optimal. Because one objective value increases
between jobs and housing for reducing trips and traffic must be at the expense of the decrease of at least one other
congestion. We calculate the number of jobs provided by objective value. So, the solution is a set of non-inferior ones,
commercial cells and the number of jobs required in station that is, the Pareto optimal solution set.
area to get the job-housing rate. This study uses χ to In this model, the decision variables include land-use type
represent the employment coefficient of the residents in study X ijk and density rijk . The objectives and constraints are the
scope. functions of X ijk and rijk . The algorithm is designed as two
phases. In the first phase, the land-use type is optimized.
1 Then the optimal density of each cell is confirmed in the
∑S ⋅ X ij ⋅ rij / ϕ
1 1
second phase according to the optimal results determined in
JRRl ≤
(i , j )
≤ JRRu , ∀(i, j ) ∈ Z (9) the first phase. The main procedure of the approach is briefly
2
χ ⋅ ∑S ⋅ X ij ⋅ rij / ϕ
2 2 illustrated in Fig. 5.
(i , j ) Start
∑S ⋅ X
(i , j )
3
ij ⋅ rij3
Calculating fitness
NO
Stop
εl ≤ ≤ ε u , ∀(i, j ) ∈ Z (10)
S ⋅ X ijk ⋅ rijk G>GEN
YES
∑∑
( i , j ) k =1,2 ϕk Selection
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
them into several subpopulations according to the number of is shown in Fig. 6(b). It shows the land-use type in study
objectives. Each subpopulation corresponds to one objective. scope includes residential, green space, transportation and
Second, calculate the objective value and fitness value of undeveloped. Those cells marked transportation represent
each chromosome by using a rank-based fitness method. roads and metro line. The green space is a park for
Select the chromosomes which have higher fitness to form surrounding residents. Almost half of the total cells are not
the new subgroups. Third, integrate all the newly generated developed, which contributes to implement transit-oriented
subpopulations into a new complete group. And then land planning around the URT station.
generate a new population by using the crossover and 1) PHASE ONE: OPTIMAL LAND-USE
mutation methods. Repeat the steps, the optimal solution According to the present land-use, the land cell matrix is
gradually comes to be stable with frequent iterations. Finally, coded as shown in Fig. 7(a). It shows the number of
the solution set is a group of Pareto feasible solutions. undeveloped cells coded by 0 is 390. Thus, the length of one
The second phase aims to decide the density based on the chromosome is 390 × 3 = 1,170.
optimal land-use type matrix. Assume that the densities of In this phase, the genetic algorithm toolbox based on
cells which have the same land-use type and zone are MATLAB software developed by Sheffield University is
identical. And the density of the cells with the same land-use used to get Pareto solutions. The MATLAB program of
type is conformed to gradient development. In this phase, the commercial type is compiled with five objectives of land-use
density rijk is the decision variable. The objective of this type and constraints.
phase is to maximize the urban rail transit ridership. So, the After 2,000 generations of iterations, the optimal value of
model turns to be a single-objective linear optimization objective one converges at 944th generation to -2.87×107, as
model. This study uses MATLAB linear programming tool shown in Fig. 8(a), i.e., the maximum of the average land cell
to solve the linear programming model. price is CNY 2.87×107 Yuan; objective two converges at
380th generation to -2.4×104, as shown in Fig. 8(b), i.e., the
V. CASE STUDY maximum of urban rail transit ridership is 2.4×104 persons in
The model are tested by using the Sidaoqiao station of peak hour; objective three converges at 456th generation to -
Beijing metro line S1 as a commercial-dominating type 1,051, as shown in Fig. 8(c), i.e., the maximum of total land-
example and Pingxifu station of Beijing metro line No.8 as a use compactness of commercial land cells in the intermediate
residential-dominating type example. zone and the internal zone is 1,051; objective four converges
at 1,488th generation to -1,757, as shown in Fig. 8(d), i.e., the
A. COMMERCIAL-DOMINATING TYPE: SIDAOQIAO maximum of land-use compactness of residential land cells
STATION
in study scope is 1,757; objective five converges at 1,067th
Sidaoqiao station is a station of Beijing metro line S1, which
generation to 314, as shown in Fig. 8(e), i.e., the minimum of
is located in the Mentougou District. This station gives
the average distance between land cells and station center is
priority to develop commerce and business.
314m. The result shows objectives one, four and five
This study investigates the present land-use within the
converge slowly. And all objectives converge before the
station study scope based on the Beijing geography
2,000th generation. Pareto solutions are acceptable by testing
information system. Google satellite map around the station
the rationality of results, shown in Table. 3.
is shown in Fig. 6(a). The land cell diagram of the study area
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
(a) Land cell matrix of current status (b) Land cell matrix of solution F
FIGURE 7. Land cell matrix of Sidaoqiao Station.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
TABLE 5. Comparison between the optimal plan and current plan of Sidaoqiao Station.
Proportion Commercial land Residential land Public land
Optimal plan 37.13% 54.21% 8.66%
Current plan 85.43% 13.33% 1.24%
Difference -48.30% +40.88% +7.42%
Five alternative Pareto solutions (A to E) are the 3) OPTIMIZATION RESULTS ANALYSIS
approximated optimal solutions for each objective. Solution It shows significant differences between the optimal
A and D show the relations between land value with solution and the current plan, as shown in Table. 5. The
residential compactness. Solution B and C show the relations proportion of commercial lands has decreased by 48.30%,
between ridership with commercial compactness. The and the proportion of residential lands and public lands has
increase of residential compactness can increase total land increased by 40.88% and 7.42%.
value because of more residential cells. Similarly, the The reasons can be summarized as follows. First, too high
increases of commercial compactness can increase the total development capacity is not suitable for the peripheral
ridership because of more commercial cells. However, stations. Too many commercial cells will cause low average
commercial and residential compactness have fewer conflicts. development density and turn to a waste of land resources.
Thus, conflicts occur between land values and ridership as Besides, it will aggravate the tidal passenger flow of urban
well. Solution A to D are two extremes. Although the rail transit and increase pressure on urban rail transit
accessibility of solution E achieves the optimum value operation. The constraint of keeping a balance of jobs and
313.79, the land value and ridership are at a low-level. housing in station area also ensure a reasonable proportion
Solution F is a satisfactory solution generated by the model between commercial lands and residential lands.
to find a balance between the five objectives. The land value In the optimal plan, the densities of all kinds of land-uses
is CNY 2.7337×107 Yuan and the ridership is 17,285 persons are at a high-level. The development capacity reaches
in peak hour, both are in a reasonable range. The objective of 95.34%, which almost achieves the maximum development
commercial and residential compactness is 924 and 1,261, limit based on land-use type solution gets in phase one.
both are at a high compact level. The objective of Meanwhile, the job-housing balance rate is 1.20 and it’s in a
accessibility is 338.72m, which means strong commercial reasonable range. The per capita public land area is 2.223
attraction to people. The land-use planning of solution F is m2/person, higher than the lower bound of 2.218 m2/person.
shown in Fig. 7(b).
2) PHASE TWO: OPTIMAL DENSITY
The phase two aims to decide 9 density variables of cells
with distinct land-use type and zone. The result of land-use
type getting from phase one shows the number of
commercial type cells in the external zone and residential and
public type cells in the internal zone is 0. Therefore, the
phase two only need to decide the remaining 6 variables,
including commercial land density in the internal and
intermediate zone, the residential and public land density in
the intermediate and the external zone. The program is
compiled with the objective of maximizing the ridership and
constraints which ensure the job-housing balance and enough
per-capita public lands. After the calculation, there is an
optimal solution for this model. The maximum rail transit
ridership of Sidaoqiao station is 16,670 persons in peak hour.
The heat map of the optimal development density of each
land cell is shown in Fig. 9. Putting it all together, the
FIGURE 9. Optimization result of land development density of
optimized results of phase one and two are listed in Table. 4. Sidaoqiao Station.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
B. RESIDENTIAL-DOMINATING TYPE: PINGXIFU The top-head platform of the train depot has been planned as
STATION residential type benefits from the innovation of hierarchical
Pingxifu station is a station of Beijing metro line No.8, which design in stereoscopic space.
is located in the Changping District. This station is located in
1) PHASE ONE: OPTIMAL LAND-USE
the large residential community of Huilongguan where gives
According to the present land-use, the land cell matrix is
priority to the real estate development.
coded as shown in Fig. 11(a). It shows the number of
Google satellite map around the station is shown in Fig.
undeveloped cells coded by 0 is 153. Thus, the length of one
10(a). The land cell diagram of the study area is shown in Fig.
chromosome is 153 × 3 = 459. The MATLAB program of
10(b). Different from the Sidaoqiao station, most cells in the
residential type is compiled with three objectives of land-use
Pingxifu station study area have already developed and most
type and constraints.
of them are residential type. Meanwhile, there has an URT
depot and maintenance base which gives service to line No.8.
(a) land cell matrix of current status (b) land cell matrix of solution D
FIGURE 11. Land cell matrix of the Pingxifu Station.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
After 1,000 generations of iterations, the optimal value of Three alternative Pareto solutions (A to D) are
objective one converges at 180th generation to -3.823×107, approximated optimal solutions for each objective. Similar to
as shown in Fig. 12(a), i.e., the maximum of the average land commercial type case, Solution A and C show the relations
cell price is CNY 3.823×107 Yuan; objective two converges between land values with residential compactness, and
at 214th generation to -2.013×104, as shown in Fig. 12(b), conflicts occur between land values and ridership. As a
i.e., the maximum of urban rail transit ridership is 2.013×104 typical residential type station, higher residential
persons in peak hour; objective three converges at 195th compactness is critical to form a community. Solution D is a
generation to -2,994, as shown in Fig. 12(c), i.e., the satisfactory solution generated by the model to balance the
maximum of land-use compactness of residential land cells three objectives. The land value is CNY 3.7263×107 Yuan
in the study scope is 2,994. The result shows all objectives and the ridership is 15,427 persons in peak hour. The
converge before the 1,000th generation. The Pareto solutions residential compactness is 2,472. All objective values are in a
are acceptable by testing the rationality of results, shown in reasonable range based on an appropriate proportion of land-
Table. 6. use type. The land-use planning of solution D is shown in Fig.
11(b).
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
3) OPTIMIZATION RESULTS ANALYSIS Because the commercial lands account for a larger
Different with Sidaoqiao station, most of the lands in percentage in the newly developed lands, it provides a better
Pingxifu station’s study scope have already been developed. service to local residents and increases the ridership.
Thus, comparing the optimal solution with the current status In the optimal plan, the density of all kinds of lands
quo, there are little differences between the two plans, as achieves the maximum development limit on the basis of
shown in Table. 8. The proportion of residential lands and land-use type solution got in phase one. Meanwhile, the per
public lands have decreased by 3.66% and 1.43%, and the capita public land area is 2.68 m2/person. It means a better
proportion of commercial lands has increased by 5.09%. level compared with the lower bound of 1.80 m2/person.
TABLE 7. Land-use optimized solution of Pingxifu Station
TABLE 8. Comparison between the optimal plan and current plan of Pingxifu station
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047207, IEEE Access
X. Y. Huang et al.: Preparation of Papers for IEEE Access (Dec 2020)
Proc. IEEE Int. Conf. Transp. Inf. Saf., Liverpool, United Kingdom, [28] K. Cao, B. Huang, S. Wang, H. Lin, “Sustainable land use optimization
2019, pp. 723-729. using boundary-based fast genetic algorithm,” Comput. Environ. Urban.
[6] H. He, L. Guo, “The study of transit-oriented urban planning methods,” Syst., vol. 36, no. 3, pp. 257-269. May. 2012.
in Proc. IEEE Int. Conf. Mech. Autom. Control Eng., Wuhan, 2010, pp. [29] P. Dutta, E. Arnaud, E. Prados and M. Saujot, "Calibration of an
4945-4948. Integrated Land-Use and Transportation Model Using Maximum-
[7] W. M. Wey, H. Zhang, Y. J. Chang, “Alternative transit-oriented Likelihood Estimation," IEEE Trans. Comput., vol. 63, no. 1, pp. 167-
development evaluation in sustainable built environment planning,” 178, Jan. 2014.
Habitat Int., vol. 55, pp. 109-123. Jul. 2016. [30] M. Q. Zou, X. Lin, C. Mao, K. Zhang, and M. Li, “Review on the
[8] P. Calthorpe, The Next American Metropolis, Ecology, Community, Theory and Planning Principle of Transit-Oriented Development,” in
and the American Dream. New York, NY, USA: Princeton Proc. COTA Int. Conf. Transp. Prof., 2014, pp. 3532-3547.
architectural press, 1993. [31] The People’s Gov. Beijing Municipality. “Beijing Standard land price
[9] R. Cervero, K. M. Kockelman, “Travel demand and the 3Ds: density, update results,” Aug. 2014. [Online]. Available:
diversity and design,” Transport. Res. D, Transport Envir., vol. 2, no. 3, http://www.beijing.gov.cn/zhengce/zfwj/zfwj/szfwj/201905/t20190523
pp. 199-219. Sep. 1997. _72661.html
[10] R. Ewing, R. Cervero, “Travel and the built environment: A synthesis,” [32] Beijing Transport Ins. “Beijing Transport Development Annual
Transport. Res. Rec., vol. 1780, no. 1, pp. 87-114. Jan. 2001. Report,” Jul. 2020. [Online]. Available: http://www.bjtrc.org.cn/List/
[11] R. Ewing, M. Greenwald, M. Zhang, J. Walters, M. Feldman, R. index/cid/7.html
Cervero, and J. Thomas, “Measuring the impact of urban form and [33] Beijing Municipal Bu. Stats. “Beijing Statistical Yearbook,” Sep. 2019.
transit access on mixed use site trip generation rates - Portland pilot [Online]. Available: http://nj.tjj.beijing.gov.cn/nj/main/2019-tjnj/zk/
study,” US EPA, Washington, DC, USA, 2009. indexch.htm
[12] C. Ding, D. Wang, C. Liu, Y. Zhang, and J. Yang, “Exploring the [34] M. Chen, J. Lu and F. Chen, “Research on the steady operation of
influence of built environment on travel mode choice considering the public transit system based on the mechanism of the interaction
mediating effects of car ownership and travel distance,” Transp. Res. A, between land use and public transit system,” in Proc. IEEE Int. Conf.
Policy Pract., vol. 100, pp. 65-80. Jun. 2017. Electr. Technol. Civ. Eng., Lushan, 2011, pp. 6820-6824.
[13] Schweitzer, Lisa. “A Review of "Suburban Nation: The Rise of Sprawl
and the Decline of the American Dream (10th Anniversary Edition)",” XIAOYU HUANG received the M.Sc. degree
J. Am. Plann. Assoc., vol. 77, no. 3, pp. 291-292. Jul. 2011. from Beijing Jiaotong University, Beijing, China,
[14] P. Calthorpe, W. Fulton, Regional City: Planning for the End of Sprawl. in 2017. He is currently pursuing the Ph.D.
Washington, D.C., USA: Island Press, 2001. degree with the Beijing Jiaotong University,
[15] J. J. Lin, T. Y. Shin, “Does Transit-Oriented Development Affect Beijing, China.
Metro Ridership? Evidence from Taipei, Taiwan,” Transport. Res. Rec., His research interests include transit-oriented
vol. 2063, no. 1, pp. 149-158. Jan. 2008. development planning, train scheduling,
[16] H. Sung, J. Oh, “Transit-oriented development in a high-density city: organization of train operations and stopping
Identifying its association with transit ridership in Seoul, Korea,” Cities. pattern optimization for urban rail transit.
vol. 28, no. 1, pp. 70-82. Feb. 2011.
[17] J. Y. Li, H. Huang, “Effects of transit-oriented development (TOD) on
housing prices: A case study in Wuhan, China,” Res. Transp. Econ.,
vol. 80, May. 2020. Art. no. 100813. QINGHUAI LIANG received the Ph.D. degree
[18] P. Q. Gu, D. X. He, Y. L. Chen, Z. P. Christopher, and Y. Jiang, from Dalian University of Technology, Dalian,
“Transit-oriented development and air quality in Chinese cities: A city- China, in 1995.
level examination,” Transport. Res. D, Transport Envir., vol. 68, pp. He is currently a Professor of engineering, and
10-25. Mar. 2019. a Ph.D. Supervisor with the Beijing Jiaotong
[19] J. J. Lin, C. N. Li, “A grey programming model for regional transit- University.
oriented development planning,” Pap. Reg. Sci., vol. 87, no. 1, pp. 119- His research interests include urban rail transit
138, Mar. 2008. network planning and optimization analysis of
[20] X. L. Ma, X. Chen, X. P. Li, C. Ding, and Y. H. Wang, “Sustainable civil engineering.
station-level planning: An integrated transport and land use design
model for transit-oriented development,” J. Clean. Prod., vol. 170, pp.
1052-1063. Jan. 2018.
[21] G. W. Lyu, L. Bertolini, K. Pfeffer, “Developing a TOD typology for ZUHUI FENG received the M.Sc. degree from
Beijing metro station areas,” J. Transp. Geogr., vol. 55, pp. 40-50. Jul. Beijing Jiaotong University, Beijing, China, in
2016. 2019.
[22] A. Sahu, “A methodology to modify land uses in a transit oriented His research interests include transit-oriented
development scenario,” J. Environ. Manage., vol. 213, pp. 467-477. development planning, and optimization analysis
May. 2018. of civil engineering.
[23] H. M. Taki, M. M. Hassan Maatouk and M. Z. Lubis, “Spatial Model
of TOD in JMR's Master Plan,” in Proc. IEEE Int. Conf. Appl. Eng.,
Batam, 2018, pp. 1-6.
[24] A. Ibraeva, G. H. De Almeida Correia, C. Silva, A. P. Antunes,
“Transit-oriented development: A review of research achievements and
challenges,” Transp. Res. A, Policy Pract., vol. 132, pp. 110-130. Feb.
2020. SHUSHAN CHAI received the Ph.D. degree
[25] J. J. Lin, C. C. Gau, “A TOD planning model to review the regulation from Beijing Jiaotong University, Beijing, China,
of allowable development densities around subway stations,” Land Use in 2020.
Policy., vol. 23, no. 3, pp. 353-360. Jul. 2006. His research interests include urban rail transit
[26] Y. Li, H. L. Guo, H. Li, G. H. Xu, Z. R. Wang, and C. W. Kong, network planning and organization of network
“Transit-oriented land planning model considering sustainability of operations.
mass rail transit,” J. Urban Plann. Dev., vol. 136, no. 3, pp. 243-248.
Sep. 2010.
[27] T. J. Stewart, R. Janssen, M. V. Herwijnen, “A genetic algorithm
approach to multi-objective land use planning,” Comput. Oper. Res.,
vol. 31, no. 14, pp. 2293-2313. Dec. 2004.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/