Transportation Planning and Technology

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/gtpt20

Optimal service zone and headways for flexible-

route bus services for multiple periods

Myungseob (Edward) Kim & Austin Roche

To cite this article: Myungseob (Edward) Kim & Austin Roche (2021) Optimal service zone
and headways for flexible-route bus services for multiple periods, Transportation Planning and
Technology, 44:2, 194-207, DOI: 10.1080/03081060.2020.1868086

To link to this article: https://doi.org/10.1080/03081060.2020.1868086

Published online: 04 Jan 2021.

Submit your article to this journal

Article views: 54

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=gtpt20
TRANSPORTATION PLANNING AND TECHNOLOGY
2021, VOL. 44, NO. 2, 194–207
https://doi.org/10.1080/03081060.2020.1868086

Optimal service zone and headways for flexible-route bus

services for multiple periods
Myungseob (Edward) Kim and Austin Roche
Department of Civil and Environmental Engineering, Western New England University, Springfield, MA, USA

ABSTRACT ARTICLE HISTORY

Where private ridesharing companies such as Uber and Lyft are Received 4 November 2019
transforming the transit sector by making flexible transit cheaper Accepted 28 October 2020
and more readily available than ever before, it may be beneficial
KEYWORDS
for public transportation agencies to adapt to the changing Public transportation;
marketplace and provide flexible bus services in low to mid- flexible bus services;
demand density areas such as suburban or rural areas. This paper headways; service area;
proposes an optimization model in which the service area for the analytic solution
flexible bus operations is jointly optimized with headways with
time-dependent demand densities. The closed-form solution that
minimizes the average costs of user and operator is obtained
with analytic optimization. The model presented in this paper
may be a helpful planning resource to determine the feasibility of
flexible bus services where fixed-route transit operation is not
desirable due to the lack of ridership, especially in suburban and
rural regions. Numerical case studies and sensitivity analyses for
critical operation factors are presented.

Introduction
Bus transit systems may be categorized into two branches: (1) fixed-route bus systems,
which can also be referred to as conventional bus services, and (2) unconventional bus
services such as dial-a-ride services, ridesharing (e.g. Uber and Lyft), or flexible bus ser-
vices. Fixed-route bus operations have characteristics that the bus routes are predeter-
mined and its operating schedules are preset. Therefore, conventional bus operations
are preferred to transport high ridership demand with high service frequencies. Conven-
tional services are typically assigned to busy corridors to transport passengers with low
average cost per person. Kocur and Hendrickson (1982) investigated local bus service
design with conventional bus operations.
Flexible-route bus services, however, have flexibility in routing and schedules. Flexible-
route bus services are typically considered for many-to-one or one-to-many demand pat-
terns such as delivering passengers from the terminal to the residential areas. Wilson and
Hendrickson (1980) proposed models to analyze system performances of flexibly routed
transportation services. Stein (1978) proposed the tour length approximation model for
dial-a-ride services, and Daganzo (1984) found that the coefficient for the tour length

CONTACT Myungseob (Edward) Kim [email protected] Department of Civil and Environmental

Engineering, Western New England University, Springfield, MA 01119, USA
© 2021 Informa UK Limited, trading as Taylor & Francis Group
 TRANSPORTATION PLANNING AND TECHNOLOGY 195

approximation are 0.90 for Euclidean space and 1.15 for rectilinear space, respectively.
Chang and Schonfeld (1991) compared fixed- and flexible-route bus services with the objec-
tive of minimizing the total cost that includes bus operator’s cost, user’s in-vehicle cost, and
user’s waiting cost. The user’s access cost is considered for fixed-route bus formulations and
door-to-door pick-up or drop-off is assumed for flexible-route bus services (i.e. no access
cost). They confirmed that there exists the demand threshold between conventional and
flexible bus services when the average cost per person is compared. Daganzo and Ouyang
(2019) recently explored demand-responsive transportation services, namely taxi, rideshar-
ing and dial-a-ride options. In that paper, several perspectives namely level of service,
resource reduction, and passenger capture among three service options are investigated.
Kim and Schonfeld (2012) extended Chang and Schonfeld (1991) to compare conven-
tional, flexible and variable-type bus services in which the variable-type bus operation is
that with the same size of the vehicle, the type of bus operation is switched between con-
ventional and flexible services based on the demand density. For instance, high demand
periods are served with conventional bus operations and flexible bus operation covers pas-
sengers with low demand periods. As the joint use (i.e. integration) of conventional and
flexible bus operations based on the demand thresholds may be beneficial to save the
total system cost, further studies are proposed such as mixed fleet operations for conven-
tional and flexible services (Kim and Schonfeld 2013), passenger transfers between conven-
tional and flexible bus operations (Kim and Schonfeld 2014), the analysis of maximum
system welfare for conventional and flexible services (Kim and Schonfeld 2015). Chien
and Schonfeld (1998) used conventional bus services as a bus feeder for a rail transit line.
The passengers, under flexible-route bus operations, do not have to transport them-
selves from their home to a pre-determined bus stop. This door-to-door functionality
is particularly favorable to elderly and disabled people as they face more challenges
than able-bodied people. In terms of the average cost per passenger, flexible bus services
are particularly favorable in areas with a lower demand density, in comparison to con-
ventional bus operations. In suburban areas where the ridership density is low in com-
parison to highly populated regions such as NYC or Boston, it provides a reasonable
transportation option to those who need it. As an extension from Kim et al. (2019)
which is to jointly optimize the service zone and headway for flexible-route services,
this paper presents an optimization model that minimizes the average cost per passenger
with the decision variables of the service area and headways for flexible services with mul-
tiple demand periods. The analytic solution is obtained for the optimal service area, and
the headway is obtained from the relation of the maximum allowable headway based on
the service area, vehicle size, and the demand density. The model presented in this paper
may be of interest for transit planners to evaluate the feasibility of flexible-route bus ser-
vices where frequent conventional services are not justified due to the low demand in
suburban or rural regions.

Materials and methods

Bus operation description and cost formulation
We seek to design the model in which many-to-one demand patterns are served. For
instance, passengers in rural or suburban areas are transported to the terminal (or
196 M. KIM AND A. ROCHE

Figure 1. Description of flexible bus tour.

Central Business District) in the downtown or vice versa. It is assumed that the route of
the bus operations and its schedules are flexible. Stein’s (1978) approximation is used to
estimate the length of the vehicle tour within the zone. The zone is assumed to be fairly
compact and fairly convex. In a larger system perspective, a region can be subdivided into
multiple service zones while each zone is connected to the main terminal for passenger
transfers so that many-to-many travel demand patterns within the region are covered
with multiple flexible-route bus services. The travel demand is uniformly distributed
over time and space within the zone, and the minimum number of stops within a
tour is at least five. A singular zone and the line haul to the terminal are shown in
Figure 1. The parameters and variables that are used in the formulations are presented
in Table 1.

Flexible bus formulation

In the flexible bus service formulations, three cost terms are considered, namely suppli-
er’s operating cost, user’s in-vehicle cost and user’s waiting cost. Kim, Levy, and Schon-
feld (2019) formulated the flexible-route bus service for one-time period so their
formulation is adapted in this study and are extended to cover demand variations
during multiple periods while the service zone is designed to be consistent.
The supplier’s vehicle operating cost is formulated with the fleet size N and the unit
operating cost of the vehicle c as shown in Equation (1). The unit cost of operation
for a vehicle is based on a fixed cost component and a variable cost component depen-
dent to the size of the vehicle (i.e. the number of seats) as in Table 1.
CS = N × c (1)
 TRANSPORTATION PLANNING AND TECHNOLOGY 197

Table 1. Notation and baseline values for inputs.

Symbol Variable Units Base value
A Zone size (area) Square miles –
a Parameter for bus operating cost $ per hour 30
ASC Average system cost $ per passenger –
b Parameter for bus operating cost $ per seat hour 0.3
CA Average total cost $ per passenger –
CT Total cost $ per hour –
CS Supplier cost $ per hour –
CV In-vehicle cost $ per hour –
CW Waiting cost $ per hour –
c Unit bus operating cost (= a + b * S) $ per bus hour –
DC Tour length within zone Miles –
hi Headway for time period i Hours –
i Number of time periods
J Line Haul distance Miles 10
l Load factor Dimensionless 1.0
N Fleet size Buses –
n Number of stops Number –
∅ Stein’s constant Dimensionless 1.15
Qi Demand density for time period i Passengers per square mile per hour –
q Demand for time period i Passengers per hour –
R Round trip time Hours –
S Bus capacity Seats per bus 45
TSC Total system cost $ –
u Number of passengers per stop Number of passengers 1
VX Line haul speed Miles per hour 30
VL Average local speed Miles per hour –
vv Value of in-vehicle time $ per passenger hour 12
vw Value of waiting time $ per passenger hour 15
w Waiting time Hours –
y Ratio of local speed to express speed Dimensionless 0.9

The length of the tour distance Dc within the zone is approximated by Stein (1978)’s
formula in Equation (2). The tour distance is the function of the number of stops in the
tour n and the size of the zone A. The coefficient ∅ is 1.15 for rectilinear space and 0.9 for
Euclidean space (Daganzo 1984). This paper assumes the rectilinear space within the
zone:
√




Dc = ∅ nA (2)
The number of stops during the tour n is found from the relation of demand density
Q, zone size, A, headway h, and the number of passengers (boarding or alighting) per
stop u:
QAh
n = (3)
u
As the actual hourly demand per zone q (in passengers/hour) is product of the
demand density Q and the zone size A, the demand q is expressed as:
q = QA (4)
Then, the tour length Dc in Equation (2) is rearranged as:







qAh
Dc = ∅ (5)
u
198 M. KIM AND A. ROCHE

The round trip time, R, is the sum of 2-directional trip time for line haul (i.e. express
segment) and trip time in the local zone. The line haul distance J, the express speed VX,
and local speed VL are used to compute the vehicle round trip time R:

2J Dc
R = + (6)
VX VL

We assume the local speed is a function of express speed VX. We denote the ratio of
local speed VL, to express speed VX, as y. Then the round travel time R is expressed in
Equation (8) as a function of express speed VX, but the local speed VL can also be
treated as an independent input parameter.

VL = y VX (7)

2J Dc
R= + (8)
VX y VX

The required fleet size N is then obtained by dividing the round travel time R by the
headway for the service h.

R
N = (9)
h
The supplier’s operating cost CS is then formulated in Equation (10).





2Jc ∅c qA
CS = + (10)
hVX y VX uh

The user’s in-vehicle cost CV is formulated by multiplying the average value of passen-
ger’s in-vehicle time vv , the actual demand q, and average travel time for passengers,
which is the half of the round travel time R, as expressed in Equation (11).

R
CV = vv q (11)
2
Equation (10) is re-written as:







vv qJ ∅vv q3 hA
CV = + (12)
Vx 2y Vx u

The average waiting time for passengers is assumed as the half of the headway. This
assumption may be valid as we develop a planning model with aggregated demand.
Thus, the user’s waiting cost CW is formulated with the average waiting time h/2, the
value of the waiting time vw , and the actual demand q.

h
CW = vw q (13)
2
 TRANSPORTATION PLANNING AND TECHNOLOGY 199

The total cost for the flexible service is the sum of operating cost, in-vehicle cost and
waiting cost as shown in Equation (14) and detailed as in Equation (15):

CT = CS + CV + CW (14)












2Jc ∅c qA vv qJ ∅vv q3 hA h
CT = + + + + vw q (15)
hVX y VX uh Vx 2y Vx u 2

The average cost per passenger CA can be found by dividing the total cost in Equation
(15) by the passenger flow q:

CT CT
CA = = (16)
QA q

Thus, the average cost for the service is:










2Jc 1 ∅c QA · A 1 vv QA · J
CA = + +
hVX QA y VX QA uh QA Vx











∅vv 1 (QA)3 hA vw QAh
+ + (17)
2yVX QA u 2QA

Equation (17) is rewritten as:











2Jc ∅c 1 vv J ∅vv A Qh vw h
CA = + + + + (18)
VX QAh y VX Qhu VX 2yVX u 2

Equation (18) is can be optimized for the service area A and the headway h for one
time period (or one time demand). As we assume the maximum allowable headway
Sl
policy hmax = is used, Equation (18) can be reduced for one decision variable
QA
(i.e. service area A) when we extend the model for multiple time periods. The average
cost for the flexible-route bus service with the maximum allowable headway policy is for-
mulated as Equation (19).









2Jc ∅c A vv J ∅vv ASl vw Sl
CA = + + + + (19)
VX Sl y VX uSl VX 2yVX u 2QA

The total cost is product of the average cost per passenger CA and the demand q, as
expressed in Equation (20) and is rearranged in Equation (21).

CT = CA × q = CA × Q × A (20)











2JcQA ∅cQ A3 vv JQA ∅vvQ A3 Sl vw Sl
CT = + + + + (21)
VX Sl y VX uSl VX 2yVX u 2
200 M. KIM AND A. ROCHE

The total cost for one time period in Equation (21) is then used to formulate the
total system-wide cost TSC that covers multiple time periods, as denoted i in Equation (22).
C
TSC = T
 









2JcQi A ∅cQi A3 vv JQi A ∅vvQi A3 Sl vw Sl
 VX Sl + y VX uSl + VX
+
2yVX u
+
2
= (22)

The average system cost ASC is obtained by the total system cost in Equation (22) by
the total number of passengers, and is expressed in Equation (23)
C C
ASC = Tq = TQ (23)
A i
The average system cost ASC is rearranged using Equations (22) and (23).
 









2JcQi A ∅cQi A3 vv JQi A ∅vvQi A3 Sl vw Sl
1  VX Sl + y VX uSl + VX
+
2yVX u
+
2
ASC = Q
A i
(24)
Equation (24) is simplified as Equation (25), and the average system cost ASC has one
unknown variable, which is the service area A. One the system-wide decision value of A is
determined, the headway for each time period can be found by the maximum allowable
Sl
headway policy hmax = .
QA









2Jc ∅c A vv J ∅vv ASl vw Sl
ASC = + + + +  (25)
VX Sl y VX uSl VX 2yVX u 2A( Q )

Results
Analytical solution
In this paper, we seek to find the analytic solution as it can provide the closed form sol-
utions and the insight for relations among design parameters. As we develop a macro
level planning model, it also makes sense to have an analytic solution for aggregated
ridership demand.
To find the optimal value of service area A, the first derivative of the average
system cost in Equation (25) with respect to service area A is set to equal to zero.
The parameter n is denoted as the total number of demand variations (or number
of time periods).





∂ASC ∅c ∅vv Sl vw S ln
= √





+ +  (26)
∂A 2y VX uSlA 4yVX uA 2A2 ( Q )
 TRANSPORTATION PLANNING AND TECHNOLOGY 201

By solving Equation (26) we find the optimal zone size A* for the flexible-route bus
operations, as expressed in Equation (27).
 √

2/3
2nv y V Sl u
A∗ =
w X
Q (27)
∅( )(2c + vv Sl)

It is essential to check whether the optimal solution of service area A* yields the glob-
ally minimum for the objective function in Equation (25). The second-order derivative
for the service area A is shown in Equation(28) and it should be positive for all values
of A to guarantee the global minimum.






∂2 ASC ∅c ∅vv Sl vw S ln
= − √






− +  . 0 (28)
∂A2 4y VX uSlA 3 8yV X uA 3
A3 ( Q )
By rearranging Equation (28), we obtain the necessary condition for the service area A
and is the maximum possible value for A. Thus, we denote it as Amax :
 √

2/3
∗ 8nvw yVX Sl u
A , Amax =  (29)
∅( Q )(2c + vv Sl)

To ensure the optimal service area A * is always less than its maximum value as shown
in Equation (29), we divide Equation (29) by Equation (27) and the relation is expressed
in Equation (30):
 √

2/3
8nvw yVX Sl u

Amax ∅( Q )(2c + vv Sl)
=  √

2/3 (30)
A∗ 2nvw y VX Sl u

∅( Q )(2c + vv Sl)
√



We find that Equation (30) is simplified as 3 16 = 2.52. Therefore, it is confirmed that the
analytical solution obtained in Equation (27) yields the globally minimum solution.

Baseline numerical analysis

Input values
As a baseline analysis, we consider multiple time periods for the analysis. It is assumed
that the demand for one day of the service is categorized into 10 demand densities, and is
sorted with the decreasing order, as shown in Table 2. Other input parameters and values
applied are presented in Table 1.

Table 2. Demand inputs for multiple time periods.

Time period Demand density (passengers/sq. mile) Time period Demand density (passengers/sq. mile)
1 20 6 10
2 18 7 8
3 16 8 6
4 14 9 4
5 12 10 2
202 M. KIM AND A. ROCHE

Table 3. Baseline case results.

Time Demand density Service area Headway Average cost Average system cost
period (passengers/sq. mile) (sq. mile) (h) ($/passenger) ($/passenger)
1 20 9.83 0.23 12.60 14.01
2 18 0.25 12.79
3 16 0.29 13.03
4 14 0.33 13.34
5 12 0.38 13.75
6 10 0.46 14.32
7 8 0.57 15.18
8 6 0.76 16.61
9 4 1.14 19.47
10 2 2.29 28.05

Baseline results
For a time period with a demand density of two passengers per square mile, the average
cost is the highest among the average costs shown in Table 3. As the demand density
increases, which means increasing ridership, the average system cost per passenger dras-
tically decreases. When the demand density increases from two to four passengers per sq.
mile, the average cost per passenger decreases by 30.1% from $28.05 to $19.47. As the
demand density increases further, the cost per passenger continues to decrease yet the
rate slows down. Demand density increase from four to eight and eight to 16 passen-
gers/sq. mile yield 22.0% and 14.2% decreases in the average cost per passenger, respect-
ively. The data are shown graphically in Figure 2 and are summarized in Table 3.
The headways for each time period are found based on the maximum allowable
headway policy. As the demand density increase, the required fleet size N increases as
well, and as the result, the vehicles operate more frequently, which means short head-
ways. The exponential decay pattern for headways is displayed in Figure 3.

Figure 2. System-wide average cost and average cost of trip for demand density variations.
 TRANSPORTATION PLANNING AND TECHNOLOGY 203

Figure 3. Headway variations over demand densities.

With the closed form solution is shown in Equation (27), the optimal service area A* is
found as 9.83 sq. miles, and headway decisions are found for each time period, respect-
ively, as shown in Table 3. The headway shows variations between 0.23 and 2.29 h. As we
evaluate the headway decisions for various demand density variations, it is noted that
when the demand density is very low such as two passengers per sq. mile, the
headway is determined as 2.29 h. It is possible that the optimized headway is longer
than the round travel time R. If the transit agency plans to provide at least a vehicle oper-
ating within the time of the round travel, it is necessary to adjust the headway to be less
than or equal to the round travel time R. However, it is noted the numerical example pre-
sented in this paper shows how demand variations including very low demand density
affect the headway policy decisions. In practical operations, the transit agencies typically
set the maximum headway policy, which as an example can be less than an hour (i.e.
means that the bus service is provided at least every hour). For low demand densities
namely two to four passengers per sq. mile, the resulting headway is over an hour so
it is assumed that the demand is uniformly generated within the zone while the
vehicle is being held at the terminal due to the low demand.

Sensitivity analysis
In the sensitivity analysis, we explore how the variation of critical variables affects the
system cost of the flexible bus services. The input values for demand densities and
other parameters are kept the same as the baseline case study.

Vehicle size
We change the input values of the vehicle size S, ranging from 5 seats per bus to 45
seats per vehicle. As the vehicle size increases, the area of coverage and cost per
204 M. KIM AND A. ROCHE

Figure 4. Sensitivity of service area and cost with respect to vehicle size.

passenger also increase. This is due to the capacity of the vehicle to carry more pas-
sengers on a given trip. It is noted that as vehicle size increases, the optimal value of
service area A and the total system cost increase. The sensitivity of system costs with
respect to the size of the vehicle is shown visually in Figure 4 and are summarized in
Table 4.

Number of people per stop

Increasing the number of people per stop also increases the efficiency of a flexible-route
bus operations. In this sensitivity analysis, we vary the value of u (number of people per
stop), ranging from 1.0 to 2.0. When the value of u increases, the number of stops n
decreases and it results in decreasing tour distance. This causes the average cost per pas-
senger to decrease significantly. As a result, this increase in efficiency causes a slight
increase in service area for one vehicle. This increase is shown in the analytical solution.
√


The analytical solution shows that the service area is proportional to 3 u. The data are
shown visually in Figure 5 and is summarized in Table 5.
As the service area A increases with the number of people per stop, this inevitably
increase the total system cost. This is because, assuming uniform demand density, an
increase in service zone results in more ridership generated and, therefore, it increases
the total system cost. Although the total cost is rising in this scenario, it is important
to remember that an increase in number of riders per stop causes a decrease in cost

Table 4. Sensitivity of system cost over vehicle size.

Vehicle size (seats/bus) Service area (sq. mile) Average system cost ($/passenger) Total system cost ($/day)
5 3.24 $11.36 $4,043.57
10 4.91 $17.85 $9,643.52
15 6.07 $25.04 $16,710.57
20 6.96 $32.91 $25,190.90
25 7.69 $41.41 $35,047.89
30 8.33 $50.51 $46,254.72
35 8.88 $60.18 $58,790.73
40 9.38 $70.41 $72,639.46
45 9.83 $81.17 $87,787.38
 TRANSPORTATION PLANNING AND TECHNOLOGY 205

Figure 5. Sensitivity of service area and cost with respect to number of people per stop.

Table 5. Sensitivity of system cost over the number of people per stop.
Number of people per stop Service area (sq. mile) Average system cost ($/passenger) Total system cost ($/day)
1.00 9.83 $14.01 $15,147.93
1.10 10.15 $13.71 $15,310.07
1.20 10.45 $13.45 $15,462.66
1.30 10.73 $13.22 $15,606.99
1.40 11.00 $13.01 $15,744.09
1.50 11.25 $12.82 $15,874.82
1.60 11.50 $12.65 $15,999.85
1.70 11.73 $12.49 $16,119.78
1.80 11.96 $12.34 $16,235.09
1.90 12.18 $12.20 $16,346.21
2.00 12.39 $12.08 $16,453.49

Figure 6. Sensitivity of service area and total cost with respect to number of people per stop.

per passenger. Doubling the number of passengers per stop from one to two yields an
11.6% decrease in average cost per passenger. The data are presented in Figure 6 and
Table 5.
206 M. KIM AND A. ROCHE

Conclusions
This paper has formulated a solution for flexible-route bus operations with demand vari-
ations. The previous study by Kim, Levy, and Schonfeld (2019) has been extended to con-
sider multiple time periods. The optimal solution for the service area was obtained
analytically with the assumption of the maximum allowable headway policy. With the
optimal service area, which is constant across the multiple time periods, the optimal
headway was obtained for each time period. The results of the numerical analyses
showed that a lower demand requires a system with a high average cost per passenger.
We noted the drastic fluctuations in average cost per passenger from low-demand to
high-demand time periods. As expected, for the high demand time period, the optimal
service area decreased and the headway also decreased. Sensitivity analyses found that
increasing the vehicle size or increasing number of people per stop increased the
optimal service zone area. As the number of people per stop increased, we noted the
total system cost was increasing mainly due to the increased service area generating
more ridership. However, this increase resulted in the decreased number of stops so
that the average cost per passenger decreased.
The analytic closed-form solution provides the insights between the design parameters
and decision variables. For instance, the value of waiting time positively affects the
optimal service area. The increase of the vehicle operating speed also increases the
service coverage of the vehicle as it decreases the travel time for both passengers and
vehicle operators. We also found that the sum of demand density negatively affected
the optimal service area. Regarding the global optimality of the solution, we confirmed
that the optimal value of service areas (shown in Equation (27)) is always less than its
maximum value (presented in Equation (29)). The analytical solution and cost formu-
lations for flexible-route bus services may be used as a useful guideline to examine an
alternative transit operation where the demand is not sufficiently high to justify frequent
services using fixed-route bus operations.

Acknowledgements
This study was carried out with the support of the Western New England University Faculty Grant
Program.

Disclosure statement
No potential conflict of interest was reported by the author(s).

ORCID
Myungseob (Edward) Kim http://orcid.org/0000-0002-8433-4950

References
Chang, S K, and P Schonfeld. 1991. “Optimization Models for Comparing Conventional and
Subscription Bus Feeder Services.” Transportation Science 25 (4): 281–298
 TRANSPORTATION PLANNING AND TECHNOLOGY 207

Chien, S., and P. Schonfeld. 1998. “Joint Optimization of a Rail Transit Line and Its Feeder Bus
System.” Journal of Advanced Transportation 32 (3): 253–284.
Daganzo, C. 1984. “The Length of Tours in Zones of Different Shapes.” Transportation Research
Part B: Methodological 18 (2): 135–145.
Daganzo, C. F., and Y. Ouyang. 2019. “A General Model of Demand-Responsive Transportation
Services: From Taxi to Ridesharing to Dial-a-Ride.” Transportation Research Part B:
Methodological 126: 213–224.
Kim, M., J. Levy, and P. Schonfeld. 2019. “Optimal Zone Size and Headways for Flexible Bus
Services.” Transportation Research Part B: Methodological 130: 67–81. https://doi.org/10.
1016/j.trb.2019.10.006
Kim, M., and P. Schonfeld. 2012. “Conventional, Flexible and Variable-Type Bus Services.” Journal
of Transportation Engineering 138 (3): 1–11.
Kim, M., and P. Schonfeld. 2013. “Integrating Bus Services with Mixed Fleets.” Transportation
Research Part B: Methodological 55: 227–244.
Kim, M., and P. Schonfeld. 2014. “Integration of Conventional and Flexible Bus Services with
Timed Transfers.” Transportation Research Part B: Methodological 68: 76–97.
Kim, M., and P. Schonfeld. 2015. “Maximizing Net Benefits for Conventional and Flexible Bus
Services.” Transportation Research Part A: Policy and Practice 80: 116–133.
Kocur, G., and C. Hendrickson. 1982. “Design of Local Bus Service with Demand Equilibration.”
Transportation Science 16 (2): 149–170.
Stein, D. 1978. “Scheduling Dial-a-Ride Transportation Systems.” Transportation Science 12 (3):
232–249.
Wilson, N. H. M., and C. Hendrickson. 1980. “Performance Models of Flexibly Routed
Transportation Services.” Transportation Research Part B: Methodological 14 (1–2): 67–78.