S2 201700035 Fierek Sawicki Ver2
S2 201700035 Fierek Sawicki Ver2
S2 201700035 Fierek Sawicki Ver2
1 Introduction
Public transport offers many advantages over individual ways of moving by private
transport means, including: lower transportation costs, lower utilization of space per
person travelled, lower energy consumption and lower pollution. Public transport
increases overall mobility of the local community, especially of its part that does not
possess private cars. It also improves accessibility to different destinations, including:
places of employment, business activities centres, points of interests and/or recrea-
tional areas. For these reasons, local authorities especially in metropolitan areas are
vividly interested in providing satisfactory level of public transport services, resulting
in the overall increase of the urban life standard. Thus, designing of the mass transit
system is crucial issue in urban management.
Public transport system design may be considered from different points of view
depending of managing entity, i.e.:
adfa, p. 1, 2011.
© Springer-Verlag Berlin Heidelberg 2011
planning process in public transportation for several phases such as line route plan-
ning, timetabling, and vehicle and crew scheduling.
From methodological point of view different approaches are used to solve above
mentioned problems. Lownes and Machemehl [3] proposed mixed integer model for a
single-route circulator design problem, while Szeto and Wu [4] used metaheuristic
algorithms (hybrid genetic algorithm) to solve simultaneously both the route design
and frequency setting problems. Cancela et al. [5] propose to solve the problem of
calculating the number and itinerary of bus routes and their frequencies for a public
transportation system, using mathematical programming.
In this paper, the authors consider the line route planning allied with location of
vehicle’s depots. This approach is quite novel and has not been discussed in the litera-
ture yet. The problem of depots in the transport network is considered only through
the deadheading context. Eberlein et al. [6] have considered a problem for deadhead-
ed vehicles starting their trip empty until a station to be determined, and from that
stop, these buses start their normal service until the end of their route.
The essence of combining two strategic problems of the public transport system, i.e.
lines construction and depots location, is to find a solution that guarantee the lowest
operating costs for the operator, while keeping all the required standards for passen-
gers. The nature of each decision problems is contradictory while are considered to-
gether. The problem of transport lines construction is strongly linked to the passen-
gers’ demand on movement, while depots location is dictated by searching the least
expensive land costs, which means a location of depots on the suburbs. From the eco-
nomical point of view it should be located as closed to each transport line as possible.
The combined solution of these two problems is primarily driven by the demand for
mass transit and the location of the depots closest to the concentration points of a
highly frequently utilized transport lines.
In the proposed methodology of simultaneous solving the combined lines construc-
tion and depots location problems, the authors have defined the following assump-
tions:
Fig. 1. The concept of the methodology of combined solving transport lines construction
and depots location problems (Source: Own work)
At the same time, the fleet needed to perform a transport service is deployed to the
most suitable depots. The basis for the implementation of steps 3.1-3.3 is the optimi-
zation model (see the following sub-sections).
As a result of the combined optimization of the transport lines construction and de-
pots location (step 3), the volume of supply is defined. The supply volume in step 4 is
the basis for determining the modal split, i.e. the division into private and public tran-
sport. In step 5, all the trips performed by inhabitants are assigned according to the
defined modal split, and during step 6 transport model is calibrated. If the result of
calibration is satisfactory, the calculation procedure is terminated, otherwise the pre-
vious steps (1 or 2 or 4) are performed again and the subsequent steps of the proce-
dure are repeated until the result is satisfactory.
Realization of steps 1, 2 and 4 to 6 is performed using a dedicated macro-simula-
tion tool for traffic modelling. Overview of these types of tools in the context of their
key functionalities is presented in [11]. Step 3 (i.e. 3.1-3.3) is run in an optimization
tool with a dedicated engine for linear programming problems.
( )
L I M
L = ∑ zl ∑∑ zilm ⋅ d ilp + d ilk , [PLN/24h] (3)
l =1 i =1 m =1
Constraints
The set of feasible solutions is constructed with the following formulations:
• a total capacity of the fleet of vehicles and its operational frequency are fundamen-
tal parameters for serving passenger volume on arc (j,k) of i-th transport line (see
exp. 4),
• a total number of depots in transport system is limited to N (see exp. 5),
• all the vehicles in a transport system, i.e., assigned to all transport lines, have to be
assigned strictly to available depots (see exp. 6) and a maximum number of vehi-
cles available at i-th transport line is resulted from exp. (7).
I
∑λ ⋅q ⋅ x
i =1
i i ijkt ⋅ yit ≥ Pjkt ; ∀ j = 1,..., J ; k = 1,..., K ; j ≠ k ; t = 1,..., T (4)
∑z
l =1
l ≤N (5)
I L M L
∑∑∑ z
i =1 l =1 m =1
ilm ≤ ∑ zl ⋅ ql
l =1
(6)
⎡⎛ J K ⎞ ⎤
max rit = max⎢⎜⎜ yit ∑∑ xijkt ⋅ 2d jk ⎟⎟ / vit ⎥, ∀ i = 1,..., I (7)
1<t ≤T 1<t ≤T
⎢⎝ j =1 k =1 ⎠ ⎥
Volume Volume
Nodes djk vjk Nodes djk vjk
[pas./24h] [pas./24h]
[km] [km/h] [km] [km/h]
j k Pjkt Pkjt j k Pjkt Pkjt
1 2 95 35 .11 30 140 133 1143 720 1.61 11
1 6 811 889 .80 17 140 139 2868 4439 3.31 7
1 115 816 799 .58 17 140 142 2907 1662 .98 7
2 6 851 931 .74 17 141 133 480 645 .80 21
3 62 1620 1675 .87 11 141 138 1119 1021 .40 17
… … … … … … … … … … … …
70 71 614 600 .20 18 207 52 1638 1731 .32 11
70 116 913 894 .17 16 207 60 639 639 .11 18
70 200 280 313 .40 26 208 207 999 1091 .06 17
71 67 473 400 .30 22 209 204 891 713 .17 17
71 197 141 200 .35 29 209 208 999 1091 .17 17
72 2 42 173 1.02 52 209 210 201 286 .20 25
72 5 2207 2260 .24 8 210 205 536 333 .15 20
72 206 2242 2057 .28 8 210 206 73 360 .06 48
During step 3 an optimisation model has been applied and parameterised first. The
complete list of parameter values is presented in Tab. 2.
Parameter
Comments
Notation Unit Value
cijk [PLN/km] 5.2 unit cost in medium sized transport operator
qi [pas.] 105 a total capacity of exemplary city bus – Solaris Urbino12
λi [-] .75 a passenger comfort factor, i.e. at most 75 of available
capacity should be utilised
N [-] 1 a limited number of depots in transportation system
D6
111
126
194 D1
196 129
143 144
74 167
192
161
32
162
191 127
163 31
34
145 115
1
37 6 69
38
86
10 85
107 23 66
27
22
11
51
5 30
72
D5 14 43 206
91 60
16 19 57 61
15 33 21 52
17 93
45
44 92 62
4
20
8 138
D2
149
177
171
137
176
D4
73
142
140
D3
139
Fig. 2. SimplificationOperator:
VISUM 14.00 PTV AG
of considered transportation network with possible localization of de-
temp1h.ver
pots (red
Created on:triangle:
05.04.2017D1-D6). More detail network is in the background (grey links and nodes)
1:40570
During step 3.1, a structure of the transport lines has been constructed as a result of
searching for minimal number of lines that are able to handle a significant part of the
traffic assigned to the individual arcs of the network, at the lowest cost.
As a result in Tab. 3 a set of 10 selected lines and their one-direction total length is
presented. In step 3.2 a number of vehicles required to handle the traffic assigned to
the network arcs is determined. The basis for that purpose is: velocity resulting from
traffic conditions (see Tab. 1 for each arc), capacity of the vehicle and comfort factor
(see Tab. 2) and line length (see Tab. 3).
Table 3. A set of lines, line-depot distances and vehicle allocation to each line
The resulted total number of 182 vehicles distributed to 10 lines is shown in Tab.
3. Finally, during step 3.3 all the vehicles resulted from step 3.2 are allocated to one
of the potential depot (N = 1). Based on the global optimisation model (step 3.1-3.3
are linked each other and computation is performed simultaneously) a minimum cost
objective function has been achieved: C = 20,630.0 PLN, with TP = 17,426.3 PLN
and L = 3,203.7 PLN, respectively. As a result a D4 (see Fig. 2) vehicle’s depot is
selected (l = 4) and all 182 vehicles should be served by that location.
In the step 4 (modal split) calculation of shares of trips made by public as well as
private transportation means was performed. These computations were made based on
the skim matrix consists of journey time (between particular zones). Thus the trans-
portation supply (transportation network, public transportation line routes with time-
tables etc.) and traffic conditions as well are crucial components. The result of this
step is a set of O-D matrices separate for each transportation mode. In the considered
situation, the ratio of trips made by public and private transportation is 42% to 58%
respectively. Finally, in the last step 5 (traffic assignment), allocation of demand to
supply has been carried out.
Since the first iteration assumed a traffic assignment in step 2 without modal split
structure, further iterations are necessary to determine the final line construction and
base location. As a result in Tab. 4 traffic allocation to the arcs of a considered net-
work with modal split from the first iteration is presented. In fact traffic allocation
during morning rush hour (7am-8am) is presented; it is 12.9% of 24-hour traffic.
Table 4. A traffic allocation to arcs of the network for morning rush hours (7am-8am) for
public transport only
Volume Volume
Nodes djk vjk Nodes djk vjk
[pas./h] [pas./h]
[km] [km/h] [km] [km/h]
j k Pjkt Pkjt j k Pjkt Pkjt
1 2 0 0 .11 35 140 133 47 0 1.61 35
1 6 1 47 .80 35 140 139 371 573 3.31 43
1 115 47 1 .58 45 140 142 475 475 .98 47
2 6 0 0 .74 35 141 133 0 47 .80 35
3 62 77 84 .87 44 141 138 109 100 .40 31
… … … … … … … … … … … …
70 71 0 0 .20 45 207 52 411 .32 32
70 116 0 0 .17 45 207 60 287 .11 37
70 200 0 0 .40 45 208 207 135 167 .06 29
71 67 0 0 .30 45 209 204 167 135 .17 27
71 197 0 0 .35 35 209 208 135 167 .17 29
72 2 0 0 1.02 35 209 210 0 0 .20 35
72 5 329 390 .24 32 210 205 17 0 .15 35
72 206 390 329 .28 32 210 206 0 17 .06 35
During the second iteration, as a result of application the optimization model, the
network structure is constructed with 9 (see Tab. 5) lines, i.e. line no 2 available in
Tab. 3 is rejected here.
Table 5. Line structure, line-depot distances, vehicle allocation to each line and to the depot
241
139
47
11
29
63
17
23
28
11
17
49
63
3
135
11
284
49
224
92
23
11
63
17
36
11
29
37
46
3
44
108
275
209
75
23
47
29
17
63
37
54
62
104
108
74
54
212
114
155
125
85
75
84 84
62
148
84
104
195 77
64
74
98 77
114
125
85
64
92
51 51 60
108
92
212
56
155
56
75
60
43
159 66
79
26
51
63
97
144
56
56 10
97
8
75
221
403
367
188
86
79
37
52
221
37
403
52
57
56
367
63
51
188
50 49
84
29
52
50 49
50
194 17
33
244 27
7
123 17
109
27
22
262
37
275
132
36
142
194
123
45
84
37
51
43
63
43
4
26
56
54
26
13 37
37
57
37
4
36
58
37
66
91
37
71
52
4
37
36
36
37
37
113
128
51
36 4
63
136
122
109
56
57
262
37 59 4 37
275
94 37
132
37
77
58
69
77 59 67
93
63
65
5
59
67
94
11 8
22
67
12
62
8
62
3
19
0
69
13 2
10
62 152
43
62
12
6
63
201
11
11
51
155
63
56
57
62
62
59
240 174
94 43 158
15 5 147
10
240
12
69 227
3
100
198
162
225
11 13
8
37
13
37 73
155 134
93
12 9
12
227
6
26
240
0
10
46
43
4
57
75
96
155
35
134
69
62
59
25 1
57
210
105
14
198
227
35
262
104
155
35
13
77 134
2
7
193
13
131 90 258
77
129
10 1
4
106
10
14
122
264 46
77
46
187
3
17
90 187 250
176
14 4
13
6
26 1 144
7
12
133 250
15
13122
118 4 66 271
13
1 93
131 264 206 113
29
21
120
271 185
12 9
177
269
11 3 264 89 11
6
22
8
256
57
75
8 12 9 185
96
35
66
141
213
16
43
20 3
8
58
213 118
5
11
85
118
36
16
58 47
85 72
12 2
36 16 27
6
27
132
58
295
25
21 3
281
6
11
153
85
36
92
33 122
0
198
30
10
1
122
D2
1 192
60
78
14
0 135 227
7
11
14 3
9
3
11 6 227
5
5 135
60
35
29
0
60 11
0 9
11 11 6
9 5
29
60 11 6
0 5
11 35
9
11 6 23
5
89 27
8
13 23
28
5
10 7 27
92
11
18
9
18 1
5
10
89
5
19
3
25 7
11
23
55
163
267
162
107
16 2
3
32
18 7
5
10
16 9
8
34
21 4
8
11
Fig. 3. A sample of final results - traffic volumes (blue bars and values – private transporta-
tion flows, red bars and values – public transportation flows) Legend
Links Link bar
Active Volume PrT [veh] (AP)
403
101 202
0
Passive
Volume PuT [Pers] (AP)
221
55 110
0
At the present stage of the research three potential directions of a development of the
proposed methodology can be formulated. First, it is proposed to reflect at the step 1
expected changes of transport behaviour in the long-term horizon, e.g. 5, 10 and 15
years perspective. The obtained results should allow determining the extent of the
necessary changes in both considered problems, i.e. the lines structure and the depots
location, taking into account predicted changes in mobility.
Another direction of the research is the extension of the scope of both combined
problems with regard to the fleet composition problem. It means different capacity of
transport means and sizing of each type should be additionally taking into account.
This problem has been extensively analysed by one of the author in his previous re-
search [12]. Incorporating fleet composition should allow a better matching of supply
to demand, and could directly result in better utilization of available transport capacity
while reducing fleet operating costs at the same time.
The last potential direction of development concerns the application of the meth-
odology into different types and real-life transport systems, including tram and bus
transport. As a result the universality of the methodology and the measurable benefits
of its practical application should be validated. The application of the proposed meth-
odology into the real-life transport network should also result in carrying out step 6 of
the methodology, i.e. calibration. In the problem considered in the paper a real traffic
condition is not available, thus a calibration of the traffic model is inapplicable.
References
1. Ceder A.: Public Transit Planning and Operation: Theory, Modelling And Practice. But-
terworth-Heinemann, Elsevier Ltd, Oxford (2007)
2. Schobel A.: Line planning in public transportation: models and methods. OR Spectrum,
vol. 34, 491-510 (2012)
3. Lownes N., Machemehl R.B.: Exact and heuristic methods for public transit circulator de-
sign. Transportation Research Part B. vol. 44 (2), 309-318 (2010)
4. Szeto W.Y., Wu Y.: A simultaneous bus route design and frequency setting problem for
Tin Shui Wai, Hong Kong. European Journal of Operational Research. vol. 209 (2), 141-
155 (2011)
5. Cancela H., Mauttone A., Urquhart M.E.: Mathematical programming formulations for
transit network design. Transportation Research Part B. vol. 77, 17-37 (2015)
6. Eberlein, X.J., Wilson, N.H.M., Barnhart, C., Bernstein, D.: The Real-Time Deadheading
Problem in Transit Operations Control. Transportation Research - Part B: Methodological,
vol. 32, No. 2, 77-100 (1998)
7. Fierek S., Żak J.: Planning of an integrated urban transportation system based on macro-
simulation and MCDM/A methods. Procedia - Social and Behavioral Sciences, vol. 54,
567-579 (2012)
8. Kupka, P., Sawicki, P.: Optymalizacja lokalizacji zajezdni tramwajowej w systemie komu-
nikacji miejskiej. Logistyka, vol. 2, 462-472 (2015)
9. Sawicki, P., Kupka P.: Ustalenie lokalizacji zajezdni tramwajowej w systemie komunikacji
miejskiej z zastosowaniem metody wspomagania decyzji. Urban and Regional Transport,
vol. 6, 6-11 (2016)
10. Sawicki, P., Kiciński M., Fierek S.: Selection of the most adequate trip-modelling tool for
an integrated transport planning system. The Archives of Transport, vol. 37(1), 55-66
(2016)
11. Redmer, A., Żak, J., Sawicki, P., Maciejewski, M.: Heuristic approach to fleet composition
problem. Procedia − Social and Behavioral Sciences, vol. 54 (4), 414-427 (2012)