Professional Documents
Culture Documents
Articulo 6 Kathuria2019 - Article - AReviewOfServiceReliabilityMea PDF
Articulo 6 Kathuria2019 - Article - AReviewOfServiceReliabilityMea PDF
Articulo 6 Kathuria2019 - Article - AReviewOfServiceReliabilityMea PDF
https://doi.org/10.1007/s13177-019-00195-0
Abstract
The aim of this paper is to review various public transport reliability measures including the factors responsible for causing
variability in the travel time. A four quadrant approach is used to summarize various reliability measures. These indicators use
both Intelligent Transportation System (ITS) data and the stated preference data to measure reliability. Further various supply and
demand side factors causing uncertainties in travel time are listed and discussed. In the end a brief case study on two routes of Bus
Rapid Transit System(BRTS) of Ahmedabad is reported to apply and test the reviewed travel time measures on the ITS data.
High High
Travel Average
Percepon
Time Performance
Passengers
Waing Agencies
Using Indicators
Actual Service
Time
WDI, PTI, BTI or
headway adherence
Low Low
2 Measuring Reliability of Public four contains the most important reliability measures i.e. the
Transportation Systems travel time reliability (TTR) measure. In total ten statistical range
parameters are listed in the TTR quadrant. These measures are
Figure 2 demonstrates four quadrants showing reliability mea- based upon the travel time distribution characteristics. Both re-
sures based on ITS, user perception and passenger waiting time liability and variability of travel time can be estimated using
data. Quadrant 1 illustrates headway regularity measures. In various measures as listed. The details of all the measures and
total five measures have been listed in this quadrant. The second their application by various researchers and transport authorities
quadrant contains waiting time based reliability measures. Total are studied in next sections.
6 measures are listed here, out of these most of them are based
on stated preference (SP) survey techniques. Using SP tech-
nique, utility based models were developed as reported in vari- 3 Reliability Measures Based on Waiting Time
ous literature [5–10] The third quadrant contains transfer time
reliability measure on which a limited research has been done till Waiting time at stop is an important parameter affecting
date, as of now only three measures based on transfer waiting user perception about the reliability of the system [11].
time estimation were available in literature [2]. The quadrant Ghalomi and Ziaee [12] considered average passenger
waiting time at stop as a measure of public transport Where E(U) is expected utility, ASE and ASL are
performance. Overall three types of waiting time models adherence to schedule early and adherence to schedule
were observed in the literature viz., mean delay, vari- late respectively, ‘s’ represents the cost associated with
ance delay and scheduling delay. A mean delay model adherence to schedule and PLS is probability of not
is used for PT systems because this approach focuses adhering to schedule. The models shown in Eq. (3)
only on delays related to the schedule or time table and (4) presents the scheduling cost approach to esti-
i.e. for the fixed service. This model can be in a form mate the cost associated to being early or being late
of utility equation in which utility represents a function based upon the responses of the passengers.
of expected delay in minutes “E(DM)” after the sched- Simple reliability measures like Bus Punctuality
uled arrival [6]. The functional form of this model is Indicator (BPI) and Public Performance Measure
demonstrated in Eq. (1) (PPM) were also reported in the literature. BPI is the
Percentage of buses departing within 1 min early or up
to 5 min late relative to schedule time. This measure
Utility ¼ T þ λ*EðDM Þ ð1Þ
was initially introduced and used by Department of
Where ‘T’ is the scheduled travel time in minutes, is the coeffi- Transportation United Kingdom [14]. PPM presents the
cient. The value of delay in minutes (average minute’s lateness) percentage of transit units on time. Department for
is estimated by a stated preference survey in which users tell Transport UK [15] uses PPM by assuming the transits
their willingness to pay to avoid 1 in ‘n’ probability of a certain units arriving on time if it is arriving within 10 min of
delay time. Trompet. al. [5] introduced a parameter named ex- the scheduled time for 1ong distance and 5 mints for
cess waiting time (EWT) which is defined as the difference short distances. Figure 3a, b illustrates results of various
between average waiting time (AWT) and scheduled waiting waiting time models developed in different literature for
time (SWT). If for a PT service transit units arrive irregularly different modes.
then the AWT is calculated as AWT = μ*(1 + s2/μ2)/2, where μ For the sake of comparison minutes of in vehicle travel time
represents the mean headway and s2 is headway variance [7]. (IVT) value equivalent to one minute change in Average Minute
In Variance Delay Model, the utility is represented as a func- Lateness (AML) and change in standard deviation as reported is
tion of standard deviation ‘f(s)’ of travel time [13]. Standard presented [4]. AML represents the mean of unexpected waiting
deviation is a conventional measures to estimate variability in time, for example in a study by Benwell and Black [16] it was
travel time. The model form is demonstrated at Eq. (2): observed that rail users perceive one minute change in AML as
equivalent to 12 min of IVT.
From various literatures the minutes of IVT equivalent
Utility ¼ T þ λ*f ðSÞ ð2Þ to one minute change in AML and one minute change in
standard deviation for various public transportation modes
The reliability ratio is estimated by dividing the coefficient of are presented at Fig. 2. On observing the Fig. 3(a) it can
standard deviation to coefficient of travel time. This reliability be inferred that among different modes of public transpor-
ratio can directly be used to measure performance of PT system. tation systems. Rail and Bus were showing the first two
In Scheduling Model, utility is represented as a function of highest minutes of IVT equivalent to AML. The highest
expected travel time ‘E(T)’, expected time before preferred ar- value was observed for the rail system i.e. to the range of
rival time ‘E(SDE)’, expected time after preferred arrival time 12 min. This study was carried out by Jansson and
‘E(SDL)’ and p is probability of arriving after preferred arrival Blomquist [17] in which they carried out valuation of
time [8]. The model form is expressed in Eq. (3): travel time by creating scenarios for with and without
the use of time table. In Fig. 3 (b) bus as a mode was
showing the highest value of equivalent IVT minutes for
Utility ¼ aEðTÞ þ βEðSDEÞ þ γEðSDLÞ þ θP ð3Þ one minute change in the standard deviation i.e. to the
range 2.5 to 3.0 min. This study was conducted by
Bates et al. [9] and Noland & Polak [10], further improved
MVA consultancy Ltd. [6] for both rail and bus services.
the existing utility model given in eq. 3. They suggested im-
A different kind of waiting time study was carried out by
provement in the model by adding additional parameter like
Hess et al. [18]. In this study the when the college stu-
Adherence Schedule Early (ASE) and Adherence Schedule
dents were offered a free ride for an otherwise 75 cent trip
Late (ASL) and the revised form is provided in Eq.(4)
then they were ready to wait for 5 min. Their behavior
suggested that the disutility of time spent in waiting for a
EðUÞ ¼ αEðTÞ þ βEðSDEÞ þ ΥEðSDLÞ þ θPL free ride is less than $8.5 per hour. If the same waiting
time was given to the paid trip passenger then they over-
þ βS E½ASEðth Þ þ Υs E½ASLðth Þ þ θs PLS : ð4Þ estimate this wait time by a factor of 2.
Int. J. ITS Res.
4 Reliability Measures Based on Headway the later gives an idea about the long gaps between the
Regularity vehicles.
distribution. The COV can be mathematically estimated as FHWA [43] reported a popular indicator to measure
given in Eq. 5. travel time reliability i.e. Buffer Time Index (BTI).
Equation 8 can be used to estimate BTI mathematically.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi BTI generally presents the extra buffer time (95th per-
1 D centile TT – average TT) which is required to ensure on
∑ TTd;p −TTp
D d¼1 time arrival.
COV ¼ ð5Þ
TTp
BufferIndexðBIÞ ¼ ½95thpercentileTT−½ATT=ATT ð8Þ
In Eq. 5, during the time window p, COV is the coefficient of
variation for D number of days. The acronym TTd,p is the aver-
age travel time for d days and for the time window p. TTp is the In Eq. 8 the acronym ATT is the average travel time
mean value of travel time for the time window p. A higher value and TT is the Travel time. FHWA [43] suggested an
of COV will indicate more variation in travel time and eventu- indicator that is useful for the passengers to plan their
ally less travel time reliability. Sekhar and Asakura [40] reported trip. The indicate name is Planning Time Index (PTI).
another important measure of TTR viz., planning time. The As per FHWA [43] this indicator is estimated as the
planning time is also known as the 95th percentile of travel time, ratio of the 95th percentile TT and free flow TT, later
it tells that how worst can a transit travel time delay be. The Ma et al. [35] modified this index for public transpor-
planning time as a measure is useful for the passenger informa- tation systems as expressed in Eq. 9:
tion. Mazloumi et al. [41] suggested a difference in percentile
measure i.e. the T90-T10, it is the difference of 10th percentile of
Planning Time Index ðPTIÞ
travel time from 90th percentile of travel time. A high difference
suggests high TTV and correspondingly low TTR. Lint & 95th percentile travel time
¼ ð9Þ
Zuylen [42] and Mazhloumi et al. [41] proposed a travel time average travel time
distribution based measure to estimate the TTR i.e. λ var. This
measure can be estimated as a ratio of difference of 90th and
10th percentile to the 50th percentile of travel time. Equation 6 A PTI of 1.20 will give an information to the pas-
can be used to estimate λ var, a higher value of this measure will senger that they have to plan for an additional 30 per-
present more width of the travel time distribution. centage of TT above the off-peak hour TT to ensure the
95 percentage of the on time arrival. Chepuri et al. [44]
used this indicator in their study as a ratio of planning
90th percentile TT−50th percentile TT
λvar ¼ ð6Þ time and free flow TT. Kieu et al. [45] proposed anoth-
50th percentile TT er travel time distribution based approach i.e. RTA (re-
covery time approach). RTA is a probabilistic approach
Lint et al. [42] in a study suggested a new travel time
which estimates the probability of the actual TT to be
measure i.e. λ skew. This measure can be estimated as given
higher than the scheduled TT as expressed by Eq. 10.
in Eq. 7 which gives a probability of extreme travel time. A
higher value of the measure indicates that the probability of
extreme travel time occurrence is high. Pr ðTTd:r:s ≥β þ T 50r; s Þ ð10Þ
Vehicle
Direconal Flow
Breakdown
at Intersecon
Weather
Public Transport TTV
Int. J. ITS Res.
Reliability Time Index: It is estimated as 95thpercentile TT time period thresholds δ;1 ; δ;2 are used to determine on time
minus 50th percentile TT divided by 50th percentile TT [35] arrival [46].
and this is presented in Eq. (11)
PIR ¼ P trun −tsch ϵ δ;1 ; δ;2 ð12Þ Coefficient of variation of transfer delay or of the waiting
time at the transfer points. This is a simple indicator of
Where trun is the actual running time of route during one reliability and can be used on the transfer delay data ex-
bus trip, tsch is the scheduled running time of the route and tracted by the smart card data [47]
Descriptive statistics Morning off – peak Morning peak Inter-peak Evening peak Evening off-peak
Descriptive statistics Morning off – peak Morning peak Interpeak Evening peak Evening off-peak
Transfer waiting time is a direct estimation of trans- on the roads. Conventional bus transit system which moves
fer point delays. Ceder [2] reported that transfer under mixed traffic is affected by the traffic flow in terms
waiting time usually serves as a good transfer time of increased travel time. More traffic density on these trav-
reliability indicator. el section can cause congestion at few points and hence can
Expected Transfer Waiting Time Model depends on the reduce reliability of the transit service. Delay at intersec-
arrival delays. Knopper and Muller [48] derived a peri- tion is a major impedence in a transit service, this is an
odic function of the transfer waiting time model. Goverde important uncertainty factor in case of Bus Rapid Transit
[49] developed the expected transfer waiting time model System (BRTS) where concepts of transit priority signals
as a transfer delay density function for probability of train are still not implemented. The third demand side factor is
arriving on times as presented in Eq. (13) Passenger demand which can create a variability in dwell
times at different stops along different routes hence affect-
ing the reliability of the system by affecting the total travel
time [50]. Supply side factors include both internal and
Gi tTi if p ¼ 0
f i ðpÞ ¼ ð13Þ external uncertainties. Internal are like Facility Designs,
gi p þ tTt if p > 0 traffic management schemes and vehicle breakdown and
external are like weather, accidents etc. Abdelfattah and
Khan [51] demonstrated the effect of traffic accidents on
Where gi is density function of stochastic running time Ti of
the transit travel times. Noland and Polak [10] also
the feeder train denoted by ‘i’ and Gi(T) is the corresponding
commented that accidents are major cause of congestion
delay distribution function.
in urban areas and hence affecting transit reliability.
Facility design includes the route characteristics like in
the case bus service it will be the number of bus stops,
7 Factors Affecting Public Transport Service number of intersections, location of the route including
Reliability the landuse along the routes, presence of side friction like
on street parking [52, 53]. Weather conditions like heavy
The delay in trips of Public transportation users is mainly rainfall, fog can result in increased travel times [54]. Driver
caused by factors related to infrastructure design, vehicle behaviour can impact the overall travel times of the system
breakdown etc. Various factors affecting reliability can be [36]. All the above factors can cause variability in travel
listed under supply and demand side factors as demonstrat- times in different ways like vehicle to vehicle, period to
ed in Fig. 4. The first demand side factor is the traffic flow period and day to day variability [41].
Table 4 Comparison of
scheduled TT and percentile TT of Percentile Value Morning off - peak Morning peak Inter-peak Evening peak Evening off-peak
route 1
T10 (Min) 34.53(31) 36.36(35) 36.4(31) 40.01(36) 33.54(30)
T50 (Min) 38.31 41.02 40.14 42.52 39.44
T95 (Min) 43.3 50.36 44.03 48.07 46.45
Int. J. ITS Res.
Table 5 Comparison of
scheduled TT and percentile TT of Percentile Value Morning off - peak Morning peak Inter-peak Evening peak Evening off-peak
route 2
T10 (Min) 52.22(52) 57.06(56) 55.57(52) 60.23(56) 53.45(56)
T50 (Min) 57.21 62.45 60.38 67.45 57.41
T95 (Min) 76.16 78.59 70.12 79.51 67.06
8 Case Study arrival time as presented in Table 1. From the GPS data based
excel sheets received from Ahmedabad Janmarg Limited,
8.1 BRT Route Description 2125 bus trip data was correctly retrieved. It is important to
note here that the retrieved date is for half cycle which means
The case study applies the TTR measures on two routes of Bus that it is for upstream side of both the routes i.e. from
Rapid Transit System (BRTS), Ahmedabad, India. The Maninagar to Iskon for route 1 and from Maninagar to Visat
Ahmedabad BRTS is more than 10 year old system running for route 2. The travel time data obtained from two routes can
as a hybrid BRTS system i.e. maximum of the system is a be clubbed together by taking 30 min departure time window
closed system but wherever sufficient right of way (ROW) is data together as presented in Fig. 6a, b. The data presented in
not available it operates as an open BRT system. This system Fig. 6 is two weeks of travel time data. In Fig. 6, it is easy to
is considered to be an efficient transit system when compared notice the four distinct periods. The four periods are the morn-
to other nine transit systems running in India. This system is ing-peak, inter-peak, evening-peak and evening off-peak. It
managed and operated by a special purpose vehicle (SPV) i.e. can be further observed from the Fig. 6 that the TT is ranging
Ahmedabad Janmarg Limited. from 42 min to 89 min and from 3 min to 57 min for routes
The two routes considered in the study are presented in Fig. one and two respectively. Apart from the TT plots the Fig. 6
5a, b. The route one originates from Maninagar BRTS stop also demonstrates average TT, 90th percentile TT and 10th
and terminates at the Iskon BRTS stop. Whereas the route two percentile TT.
originates from Maninagar BRTS stop and terminates at Visat
BRTS stop. The route one starts from the south eastern part of 8.3 Within a Day TTV
the Ahmedabad area and ends at the western part of the city.
On the other hand, route two starts from south western area of If the difference between the T90 and T10 plots is high it
the city and ends at city north. The total lengths of route 1 and means that the TTV for that particular time period is high.
2 are 12.1 km and 22 km respectively. The total number of bus More TTV was observed during the morning peak (08:00 to
stops intersections (signalized and un-signalized) in route 22 10:00) and evening peak-hours (16:00 to 20:00) making these
and 25 respectively, and in route 2 are 36 and 31 respectively. periods to have low travel time reliability. On the other hand
during the off-peak periods both the routes are showing high
8.2 GPS Data Collection travel time reliability. The graphs presented in Fig. 6 are easy
to understand for visually analyzing the TTR but it becomes
Ahmedabad BRTS buses are installed with GPS devices important to analyze the TTV statistically to comment about
which collects continuous bus position data at different time TTR. Therefore, Tables 2, 3, 4, and 5 presents the descriptive
intervals. The data from the GPS device is exported in XML statistics for estimating the within the day TTR at different
dialect having details like scheduled arrival time and actual periods of the day. The peak and off-peak periods of the
Friday
Thursday
Wednesday
Tuesday
Monday
Friday
Thursday
Wednesday
Tuesday
Monday
BRT system are decided by using the ridership data of the peak hours which can be confirmed by high travel time values.
system. Figure 7 illustrates the ridership data of the entire Therefore, the scheduled arrival time at the stops should con-
working day, based on this data the peak and off peak periods sider the variation in travel time so that he passengers perceive
were decided and considered in Tables 2 and 3. The mean TT a greater travel time reliability.
of route 1 varies from 36.71 min during the inter-peak period It was observed for the route 1 that trend of the mean travel
to 42.14 min during the evening peak period. And from time is parallel to the T10 and T50. Whereas, in route two the
58 min during the evening off-peak period to 68.75 min in trend of the mean TT is parallel to T10, T50 and T90. A greater
evening peak period of route two. The variation in average value of T90-T10/T50 at different periods of the day in route 1
travel time during different periods of the day is apparent. and route 2 indicates unstable travel times during these periods.
More variation in average travel time for off-peak and peak A high COV value was time is shown in the brackets which is
hours suggest that there is a delay observed at the stops during either close to or less than the 10th percentile travel time value.
Int. J. ITS Res.
This indicates that with this schedule time less than 10% of the speed was lowest during the evening peak of both the
buses will be on time. By understanding this operator should routes indicating a low performance during this period.
change the scheduled time based on the GPS data. Further TTV was high during evening off peak and and
morning off peak for route 1 and 2 respectively.
8.4 Day to Day Travel Time Variation & On comparing the schedule travel time value with the
percentile travel time, it was observed that these values
The last step in the TTR analysis of route is estimating the day are either matching with 10th percentile travel time or
to day TTV. MATLAB based TTV maps were developed in are below it. Therefore, it is strongly recommended there
the present study considering COV as a measure. The TTV is a need to revise the scheduled travel time to at least
heat maps can be observed in Fig. 8a, b. The dark areas in the match the 75th percentile travel time value during differ-
heat map indicates low TTR and the light areas in heat maps ent periods of the day.
show high TTR. The reliability maps are very easy to under- & Monochromatic maps (Fig. 7a, b) were introduced in this
stand using the scale given along with them, the operators can study to visually understand the travel time variability
use these maps to identify the unreliable periods for different within the day. These maps are created using MATLAB
days and at different times. Based on these maps further op- software and are very useful to planners and operators for
erational improvements can be suggested. improving the service. Further, for analyzing entire year’s
big data i.e. with travel time values to the range of 0.2
million, these visual maps can be very useful in clubbing
9 Conclusions the entire data in different colour bars showing average
values of COV with different darkness at different depar-
Public transport (PT) reliability is important for both passen- ture time windows.
ger and operator. Unreliable service can increase the cost of & The day to day travel time variability analysis revealed
operation and also reduce the ridership. In the present study that Monday had the highest variation in travel time. The
review of various PT reliability measures have been summa- reason for this was closely understood and it was found
rized. Few of the reviewed measures are tested on BRTS cor- that a high dwell time variation was observed on the stops
ridor of Ahmedabad using ITS data. The main findings of the during Monday. This causes variation in travel time
study are as following: resulting into low performance of the routes during this
day.
& Public transport reliability measures can be divided in four
quadrants i.e. headway regularity waiting time, transfer The case study in the present paper was limited to using only
time and travel time measures. GPS data, the review study can be extending by including var-
& These measures majorly use GPS, smart card and stated ious reliability measure based on smart card data and other sec-
preference primary survey data. The waiting time mea- ondary data sources for different public transportation systems.
sures were based on the utility based models and used
the stated preference data to build the model. Acknowledgements The authors are thankful to Ahmedabad Janmarg
Limited, India for their continuous support in data collection.
7. Osuna, E.E., Newell, G.F.: Control strategies for an idealized public 30. Rietveld, P., Bruinsma, F.R., van Vuuren, D.J.: Coping with unre-
transportation system. Transp. Sci. 6(1), 52–72 (1972) liability in public transport chains: a case study for Netherlands.
8. Hollander, Y., “Travellers’ Attitudes to Travel Time Variability: Transp. Res. A Policy Pract. 35(6), 539–559 (2001)
Intermodal and Intra- Modal Analysis”. Paper Presented at the 3rd 31. Steer Davies and Gleave (NZ) Ltd. “The Effects of Quality
International SIIV Congress – People, Land, Environment and Improvements in Public Transport. Report to Wellington Regional
Transport Infrastructure – Reliability and Development, Bari, Italy, 2005 Council, November 1990
9. Bates, J., Polak, J., Jones, P., Cook, A.: The valuation of reliability 32. Bates, J., Jones, P., Polak, J., Han, X.-L.: The investigation of punc-
for personal travel. Transportation Research E. 37(2), 191–229 tuality and reliability: re-analysis of some existing data sets. Report
(2001) submitted to rail operational research. In: Transport Studies Group.
10. Noland, R.B., Polak, J.W.: Travel time variability: a review of the- University of Westminster, London (1997)
oretical and empirical issues. Transp. Rev. 22(1), 39–54 (2002) 33. Rohr, P., Polak, J.W.: The regularity and reliability impacts of bus
11. TRB, “Transit Capacity and Quality of Service Manual”, 3rd priority measures. Proceedings. 26, (1998)
Edition. Washington, D.C., 2013 34. Currie, G., Douglas, N. J., and Kearns, I., “An Assessment of
12. Gholami, A., Ziaee, M.: Development of a performance measure- Alternative Bus Reliability Indicators”. Paper Presented at the
ment system to choose the most efficient programs, the case of the Australasian Transport Research Forum 2012 Proceedings, Perth,
Mashhad transportation system. Transp. Res. A Policy Pract. 106, Australia, 2012
261–277 (2017) 35. MA, Z., Ferreira, L., and Mesbah, M., “A framework for the devel-
13. Black, I.G., Towriss, J.G.: In: Cranfield Institute of Technology opment of bus service reliability measures”. Proceedings of
(ed.) Demand Effects of Travel Time Reliability,London: Centre Australian Transport Research Forum, Brisbane, Australia, 2013
for Logistics and Transportation (1993) 36. Strathman, J.G., Hopper, J.R.: Empirical analysis of bus transit on-
14. Department for Transport (UK), “Bus punctuality statistics 2005”. time performance. Transp. Res., Part A. 27(2), 93–100 (1993)
http://www.dft.gov.uk/pgr/statistics/datatablespublications/public/ 37. Henderson, G., Adkins, H., Kwong, P.: Subway reliability and the
buspunctuality/buspun ctualitystatisticsgb2005a, 2006. Accessed odds of getting there on time. Transp. Res. Rec. 1297, 10–13 (1993)
10 May 2018 38. Golshani, F.: System regularity and overtaking rules in bus services.
15. Department for Transport (UK), “Autumn performance report J. Oper. Res. Soc. 1993, 591–597
2007”. http://www.dft.gov.uk/about/publications/apr/ap/ 39. Kathuria, A., Parida, M., Ravi Sekhar, C.: Route performance eval-
autumnperformance07.pdf, 2007 uation of a closed bus rapid transit system using GPS data. Curr.
16. Benwell, M., Black, I.: Train service reliability on BR InterCity Sci. 112(8), 1642–1652 (2017)
services. In: Report 3: Passenger Attitudes to Lateness. Cranfield 40. Sekhar, R., Askura, Y.: Measuring travel time reliability of trans-
University, London (1985) portation system. Indian Highways. 35(11), 61–72 (2007)
17. Jansson, Kjell, and Kerstin Blomquist. "Valuation of travel time and 41. Mazloumi, E., Currie, G., Rose, G.: Using GPS data to gain insight
information-with and without use of timetable. PTRC Summer into public transport travel time variability. J. Transp. Eng. 136(7),
Annual Meeting, 22nd, 1994, University of Warwick, United 623–631 (2010)
Kingdom. 1994 42. Lint, V., J.W.C.,TU, H. and Van Zuylen, H.J., “Travel Time
18. Hess, D.B., Brown, J., Shoup, D.: Waiting for the bus. J. Public Reliability on Freeway.Proceeding of 10th World Conference on
Transp. 7(4), 4 (2004) Transport Research, (WCTR), (CDROM) Istanbul, Turkey, 2004
19. Algers, S., A. Daly, P. Kjellman and S. Widlert, “Stockholm Model 43. FHWA Report, “Travel Time Reliability: Making It There On Time,
System (SIMS): Application”, 7th World Conference of All The Time. US Department of Transportation, Federal Highway
Transportation Research, Sydney, Australia, 1995 Administration, Web Document from http://www.ops.fhwa.dot.
20. Beca Carter, “Project Evaluation Benefit Parameter Values” Beca gov/publications/tt_reliability/index.htm, 2006. Accessed 10
Carter Hollings & Ferner in Association with SDG Forsyte May 2010
Research Brown Copeland & co Report to Transfund, April 2002 44. Chepuri, A., Ramakrishnan, J., Arkatkar, S., Joshi, G., Pulugurtha,
21. Bell, D.,“Dandenong Rail Corridor Market Assessment”. S.S.: Examining travel time reliability-based performance indica-
Melbourne, a Report by Multi-Modal Transport Solutions, 2004 tors for bus routes using GPS-based bus trajectory data in India.
22. Hamilton, B.A.: “Sydney ferry business planning market research”. Journal of Transportation Engineering, Part A: Systems. 144(5),
Final report prepared for state transit, Australia. BAH Reference. 04018012 (2018)
A2526 (2001) 45. Kieu, M.L., Bhaskar, A., Chung, E.: Establishing Definitions and
23. Booz Allen Hamilton, “Sunshine Corridor Study Market Modelling Public Transport Travel Time Variability. Proceedings of
Research”. Final Report Prepared for Department of Transportation Research Board Annual Meeting, Washington, USA
Infrastructure, Victoria, Australia, 2006 (2014)
24. Douglas, N., “Sydney M2 Busway Patronage Forecasts”. Report 46. Chen, X., Yu, L., Zhang, Y.: Analyzing urban bus service reliability
Prepared for New South Wales Department of Transport, 1996 at the stop, route and network levels. Transp. Res. A. 43(8), 722–
25. Douglas Economics, “Tranz Metro Passenger Survey. Report to 734 (2009)
Tranz Metro by Douglas Economics Dated April, 2005 47. Jang, W.: Travel time transfer analysis using transit smart card data.
26. Geerts, J.-F., and Haemers, V, “Service Reliability and Transfer Journal of Transportation Research Board. 2144(1), 142–149
Convenience on the Brussels Tramway Network”. Paper (2014). https://doi.org/10.3141/2144-16
Presented at the European Transport Conference, Strasbourg, 2005 48. Knoppers, P., Muller, T.: Optimized transfer opportunities in public
27. Hensher, D., Prioni, P.: A service quality index for area-wide con- transport. Transp. Sci. 29(1), 101–105 (1995)
tract performance assessment. Journal of Transport Economics and 49. Goverde, R. M. P., “Improving Punctuality and Transfer Reliability
Policy. 36, 93–113 (2002) by Railway Timetable Optimization”. (Doctor of Philosophy), Delft
28. Kouwenhoven, Marco, et al. "New values of time and reliability in University of Technology, 1999
passenger transport in the Netherlands". Res. Transp. Econ. 47, 50. Shalaby, A., Farhan, A.: Prediction model of bus arrival and departure
2014, pp. 37–49 times using AVL and APC data. J. Public Transp. 7(1), 41–61 (2004)
29. Association of Train Operating Companies: Passenger Demand 51. Abdelfattah, A.M., Khan, A.M.: Models for predicting bus delays.
Forecasting Handbook (PDFH). Version 5, London (2009) Transp. Res. Rec. 162, 8–15 (1998)
Int. J. ITS Res.