Bus Routing Optimisation: A Case Study for the Toulouse Metropolitan

Area

Joan Burgalat

1,2 a

, Ga

el Pallares

1 b

, Myriam Foucras

3 c

and Yohan Dupuis

4 d

CESI LINEACT, Campus de Montpellier, Mauguio, France

ENSAM, Paris, France

CESI LINEACT, Campus de Toulouse, Lab

ege, France

CESI LINEACT, La D

efense, Paris, France

Keywords:

Public Transport, Transport Management and Optimisation, Sustainability, Trip Chaining, Origin-Destination

Matrix.

Abstract:

This paper investigates the efﬁciency evaluation of a public transport using an analysis of Origin-Destination

matrices (mOD). The use of a trip-chaining method on the automatically collected transport data provides

a realistic and accurate representation of trafﬁc ﬂows characterized by mOD. The introduction of a critical

walking distance and an user ﬂow at bus stop allow us to probe possible network conﬁgurations and identify

the best one in terms of service offer, ecological impact and operational cost. The conﬁgurations comparison

allows to identify the levers for the transport management. We deploy this methodology on a french case study

for the Toulouse Metropole Occitanie region. The main obtained results shows that for a walking distance

close to 1000m, the distance per day on a bus line can be optimised by 3km for a time saving close to 20%,

representing an annual gain of more than 1ton of CO

for a user loss of around 3%. These results suggest that

low-cost optimisation of a transport network is possible while maintaining a high-quality, environmentally-

friendly service offering.

1 INTRODUCTION

Public transport is a strategic and much-used mode

of transport throughout the world. It is becoming in-

creasingly necessary to promote its use and to make

this mode of transport more sustainable in order to

limit greenhouse gas emissions as CO

and pollution.

In the USA, public transport represents average 6 bil-

lion trips in 2020 and 32.87 billion total passenger

miles travelled of which 50% is for travel to work and

37% for shopping and entertainment. On the energy

front, there has been a reduction in CO

emissions

through the use of metro trains instead of private cars

and a shift in the bus ﬂeet towards hybrid buses (7%

in 2010 for 18.8% in 2020), leading to a reduction in

the use of fossil fuels (Dickens, 2023).

In Asia, domestic passenger transport is expected

to increase by 2.6% between now and 2030, ris-

https://orcid.org/0009-0006-1726-0804

https://orcid.org/0009-0002-8680-604X

https://orcid.org/0009-0002-3673-8528

https://orcid.org/0000-0002-9725-2049

ing from 16 trillion passenger kilometres in 2018 to

around 22 trillion in 2030. However, it is predicted

that CO

emissions will increase by 1.1%. Although

there is a slowdown in CO

emissions, the forecasts

suggest that the objectives of the Paris Agreement

on climate change will not be achieved (Sudhir Gota,

2023).

In EU, use of public transport increased by 70%

overall between 2000 and 2019, although there was a

decrease in some countries (Poland, Slovakia, Greece,

etc.).

It should be remembered that in the EU road

transport represents average 96% of transport sector

greenhouse gas emissions, although there has been

a decrease since 2008 which continues to accelerate

(ADEME, 2023).

Regular use of public transport has become an im-

portant part of the demand for transport networks.

Before the COVID-19 crisis in France, regular use

of public transport hovered around 70% (UTP, 2022)

with a peak in 2019 (73%) with a high proportion of

bus mode use (around 40%). Post COVID-19 use is

estimated at 53% in 2021, with a recovery to 59% in

370

Burgalat, J., Pallares, G., Foucras, M. and Dupuis, Y.

Bus Routing Optimisation: A Case Study for the Toulouse Metropolitan Area.

DOI: 10.5220/0012689100003702

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 10th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2024), pages 370-377

ISBN: 978-989-758-703-0; ISSN: 2184-495X

2022. However, 41% of French people do not use

public transport while the transport sector is the main

source of greenhouse gas emissions (MinisterForE-

cology, 2022).

Public transport is an important part of work-

ing life and is constantly growing to meet the ever-

increasing demand due to demographic growth and

urban densiﬁcation. To meet this demand, it is becom-

ing important to optimise transport networks. To do

this we need to be able to quantify and evaluate net-

work efﬁciency. Some authors have already worked

on this issue. The notion of transport network efﬁ-

ciency is a complex one: we can cite six major issues

such as cost, service, congestion, sustainability, so-

cial inclusion and accessibility, the relevance of these

issues varies depending on the observer viewpoint.

Daraio et al. propose a framework with a reading

grid weighted to take this particularity into account

(Daraio et al., 2016). For example, Hrelja explains

that there are a number of important factors involved

in assessing the effectiveness of networks. These in-

clude quality of service (waiting times, journey times,

punctuality and clarity of information). However,

these points can also depend on the geographical area

(urban or rural) and the travel objectives: for work, we

expect frequent departures and a high level of punctu-

ality; ticketing systems must be simple and integrated,

and vehicles must be comfortable (Hrelja et al., 2020).

Moreover, others factors such as bus frequency, jour-

ney time or distance travelled, are directly linked to

the network and bus routing and have an impact on

the network’s response to demand, as well as having

an impact on the environment. Network-dependent

points can be measured from data collected during bus

passages. Distance and travel time are linked to stops

imposed by the bus route to meet user demand. A

standard way of assessing demand is to estimate the

origin-destination matrix.

In the context of sustainable cities, it is impor-

tant to work on the various levers to reduce the trans-

port network’s footprint. Considering the CO

emis-

sion rate for a bus (around 900 −1000 g/veh.km) and

car (around 150−200 g/veh.km) in France (INFRAS,

2023), it takes between 5 and 7 cars to emit as much

as a bus. Looking at short-distance journeys

(< 100km), statistics show that 1.4 people on aver-

age (MinisterForEcology, 2019) occupy a car. A bus

will be more proﬁtable than a private car, if there is

a ﬂow of people at each stop in the interval [7,10].

In this work, we propose an innovative approach to

optimise a transport network from an ecological and

an economic point of view. The objective is to reduce

greenhouse gas emission while maintaining quality of

service, simply by analysing ADCS data. Daily OD

matrices are estimated using the trip chaining method

described above. The passenger ﬂow per stop (num-

ber of people using the stop to get on and off) is

then computed. The introduction of a critical walk-

ing distance and an user ﬂow at bus stop allow us to

probe possible network conﬁgurations by removing

bus stops and route modiﬁcation thanks to the Open

Source Routing Machine (OSRM) software. By con-

sidering various constraints such as the ﬂow at stops

and the acceptable walking distance threshold, we can

optimise the bus route.

• Decrease journey times and distance travelled

• Reduce greenhouse gas emissions

The paper is structured as follows: ﬁrst, we intro-

duce the experimental setup with the public transport

network and the available data. Next, the methodol-

ogy is presented with an explanation of Trip Chain-

ing Method and the algorithm we developed. Subse-

quently we present the results obtained and the dis-

cussion section offers an analysis of the results and

opportunities for further researches.

2 METHODOLOGY

2.1 Data Collection

The development and management of the network is

based on the data collected, whether ticketing data

when validating the ticket or location data collected

with the Operations Support System. General Transit

Field Speciﬁcation (GTFS) ﬁles contain network in-

formation (lines, timetables, stations, etc.) The com-

bined use of location and ticketing data provides pre-

cise spatio-temporal information on how passengers

use the network. Based on the data mentioned , we

used the mODs of a high service level bus lines to test

our method. Those mODs matrices represent regu-

lar season ticket except for unique validation ticket or

tickets with multiple validation Those trips take place

outside the school holidays and at weekends. On a

thursday in october 2019, around 24000 passengers

used this line, knowing that there 80 passages in one

way. Those data represent a high level bus line and do

not consider travel on demand transport.

2.2 mOD Reconstruction

An Origin-Destination Matrix is a matrix that de-

scribes the number of users a

i j

who boarded at ori-

gin i (row) and alighted at destination j (column).

This tool is used to represent the demand placed by

Bus Routing Optimisation: A Case Study for the Toulouse Metropolitan Area

371

Figure 1: Displacement reconstruction using TCM.

users on the transportation network. Originally con-

ducted through surveys, mODs are now established

using Automatic Data Collection Systems (ADCS)

data, which is more cost-effective and easier to imple-

ment. Networks can be divided into two categories:

closed networks, where users validate both at entry

and exit, and they have a fare dependent on the dis-

tance travelled; and open networks, where users only

validate at boarding, and the fare is independent of

the distance travelled. For closed networks, establish-

ing an mOD is straightforward. However, for open

networks, it is necessary to estimate the user’s alight-

ing stop. Various methods exist for this, including the

Trip Chaining Method (TCM) (Alsger et al., 2016;

Hora et al., 2017), which utilizes ticketing and geo-

graphical data, as well as the different boarding lo-

cations throughout the day to chain the journey and

determine alighting stops. For example, the use of

ADCS data to estimate mOD (Cui, 2006; Wang et al.,

2011; Huang et al., 2020), the identiﬁcation of fac-

tors inﬂuencing modal change (Ji et al., 2019; Li,

2016) or the evaluation of the effectiveness of com-

bining different methods and/or datasets (Feng et al.,

2018; Osorio and Punzo, 2019; Ge and Fukuda, 2016;

El Mahrsi et al., 2014). The above examples suggest

that mODs can be used to optimise public transport

networks.

In our case study, the mODs were reconstructed

using the trip chaining method (TCM). This method is

based on transport data (ticketing, Operating Support

System, etc.) and is used to reconstruct a passenger’s

journey on the network based on multiple validations

on the same day. TCM is based on various principles

that aim to describe the logic of user choices during

multiple-validation journeys. These include:

1. The alighting stop is the same stop as the next

boarding stop or the nearest stop to the next board-

ing stop.

2. The last stop of the day is the same as the ﬁrst

boarding stop or the stop closest to the ﬁrst board-

ing stop.

3. To make a connection, users agree to walk a cer-

tain distance less than a threshold distance.

This is an approximate method and those simpliﬁca-

tion due to principles implies some errors. Hussain in

(Hussain et al., 2021) explain that TCM has average

90% of success. For more information on the prin-

ciples governing TCM, please consult (Hora et al.,

2017).

In ﬁgure 1, a passenger travels on the network and

his various validations are collected at different times.

The validations correspond to the different journeys

made by the passenger on the network. One of the

principles of TCM is that the passenger gets off at the

stop closest to his new stop. We can therefore deduce

the locations of the alighting and the time using the

GTFS and reconstruct the mOD.

2.3 Network Optimisation Method

Towns are becoming denser and populated leading to

an increasing of pollution and need for mobility In

view of the current crisis in public health, the need

to improve air quality and climate change linked to

greenhouse gas emissions, we need to consider all

possible solutions to limit the impact of the players

involved in this situation. One possible solution is to

remove bus stops that are not used very often and then

calculate the shortest bus route and then reduce green-

house gas emission.

In our case study, we have chosen a line in the

Tiss

eo Collectivit

es network and the objective is to

determine which stops on this line are few used and

to see what effects the removal of these stops would

have on the network response (number of passengers

not picked up, number of km covered, etc).

To carry out this analysis, we chose to consider

two constraints: the ﬁrst is the ﬂow at the stop, i.e.

the number of people using the stop (boarding and

alighting). The second is the acceptable walking dis-

tance. We retrieve the ﬂow value N

at the stop i being

evaluated and compare it with a ﬂow threshold value

. If the ﬂow at the stop is smaller than the thresh-

old ﬂow, we remove the bus stop and the ﬂow of the

removed stop N

is put to zero. The distance threshold

will determine if users of the removed bus stop are

lost in the worst case or reported to the nearest stop.

Indeed, we calculate D

i−1,i

which is the distance be-

tween stop i and stops i − 1 and D

i,i+1

which is the

distance between stop i and i+ 1. If at least one of the

distances is smaller than the threshold distance D

then users from bus stop i are transferred to the near-

est stop implying an user walking between both bus

stops.The walking distance users to reach the near-

est stop is calculated using the OSRM application in

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

372

walking mode. Otherwise, the users are considered to

have been removed for the public transport users. For

a given distance threshold and a given ﬂow threshold,

the bus line is evaluated and if some bus stops are re-

moved, the distance travelled and the travel time by

the bus are computed with the OSRM application in

driving mode. The algorithm used to achieve this is

shown in Algorithm 1.

Data: List inter stop distance,

List stop f lows

Initialization;

for N

in List stop f lows do

if N

< N

then

The stop is removed

for D

in List inter stop distances

if D

i−1,i

< D

or D

i,i+1

< D

then

Users go to the nearest stop;

min(D

i−1,i

i,i+1

)

+ = N

;

Walking distance+ =

× min(D

i−1,i

i,i+1

);

else

Users are lost;

Loss+ = N

;

end

= 0;

else

end

Calculation of the new bus route (distance

and duration);

end

Algorithm 1: Optimisation and bus routing calculation al-

gorithm.

3 RESULTS

3.1 Experiments

We propose a case study on the public transportation

network in Toulouse, managed by the entity Tiss

Collectivit

es, which plans, organizes, and evaluates

its mobility policies. The network is made up of 2

metro lines with 27 km of track, 2 tram lines with 24.8

km of track, an urban cable car (T

eo, the longest in

France) with 3 km of track, 10 High Service Level

Buses and around a hundred bus routes covering 1494

km. The intention to develop a sustainable and re-

sponsible service includes speciﬁc facilities in the

study area, such as park-and-ride facilities and bike

Figure 2: Removed stops as a function of ﬂow threshold.

parks. Additionally, there is a Transport-on-Demand

service for an area with over a million inhabitants

(Tiss

eo, 2023).

3.2 Results

In this paper, we present results for one of the High

Service Level Bus. On the studied day, around 24000

passengers used this line, knowing that there 80 pas-

sages in one way. The line is composed of 38 bus

stops and the nominal travelled distance is 14,568 km.

Figure 2 shows the number of stops removed from the

line as a function of the ﬂow threshold values chosen.

The graph shows an increasing trend, as the higher

the ﬂow threshold, the more likely it is that a stop

will be cancelled. In order to achieve the objective

of reducing CO

, the minimum ﬂow value accepted is

7 people per stop. With 80 symmetrical passages per

day, this amounts to a minimum threshold of 560 peo-

ple using the stop per day. A graphical reading, con-

ﬁrmed by data analysis, shows that for a daily ﬂow of

560 people per acceptable stop, half of the stops on

the line will be cancelled.

Figure 3 represents the ratio of people lost on the

line as a function of the threshold ﬂow value and the

acceptable walking distance in the form of a heat ma-

trix. It can be seen that quality of service will be af-

fected for acceptable walking distances of less than

350m, whatever the ﬂow. For distances greater than

350m, we can see that the ratio of people lost is very

low, which is explained by a stop reassignment be-

cause the median inter-stop distance is 346m. There-

fore, there is a one-in-two chance of ﬁnding a consec-

utive stop within 350 metres, hence the completely

blue zone above 400 metres, as very few people are

lost.

When bus stops are removed, users have to walk

to the nearest bus stop. The ﬁgure 4 represent the

walking distance that users have to cover. This dis-

Bus Routing Optimisation: A Case Study for the Toulouse Metropolitan Area

373

Figure 3: Rate of people lost per ﬂow threshold and distance

threshold.

Figure 4: Walking distance per removed stops as a function

of distance and ﬂow threshold.

tance is a function of the ﬂow threshold and the dis-

tance threshold i.e. the distance that users are willing

to walk, however those distances do not represent an

one-way displacement and not a round trip.

For example, whatever the ﬂow threshold value,

for acceptable walking distances below 250m, users

will have to walk an extra-distance of 500m on the

network to reach a stop. From 250m to 350m, the

distance to be covered is of the order of a kilometre.

Beyond 350m, there is a fracture at a ﬂow threshold of

880 users per stop. For ﬂows above this cut-off, users

will have to walk more than 2km, but for lower ﬂows

the maximum walking distance will be close to 1km.

Figure 5 shows the bus distance travelled reduc-

tion (difference of the distance travelled by the bus

between ante and post optimisation) as a function of

ﬂow threshold. It can be seen that the curve has an in-

Figure 5: Travelled distance reduction as a function of the

ﬂow threshold.

creasing trend, but with levelling off phenomena, par-

ticularly from a ﬂow threshold equal to 560 users per

stop. In this case, the bus reduces its travel distance

by 3km per trip. For ﬂow thresholds between 560 and

1300, the gain in distance is very small because the

removal of stops does not change the bus route in this

range, the removed stops are aligned and the rerout-

ing of buses is therefore not signiﬁcant. Beyond this

range, the gain in distance is around 8km, as the re-

moval of several consecutive stops is likely to lead to

a change in the route and therefore the use of another

route.

4 DISCUSSION

The question is how the optimisation by stop deletion

affects the service quality. We have therefore plotted

on the ﬁgure 6 the distribution of users lost due to a

stop being cancelled (inﬂuence of ﬂow threshold, cf.

Figure 2) and as a function of the acceptable walking

distance.

The number of people lost is inversely propor-

tional to the acceptable walking distance. For exam-

ple, for a ﬂow threshold of 560 people per trip and

an acceptable walking distance threshold of 100m,

we lose 4651 users, whereas for a walking distance

threshold of 1200m, we lose around 601 people. For

a total attendance of 24000 people in the worst-case

scenario, this amounts to a loss of 20% of people,

whereas in the best case losses are estimated at around

3%.

The choice of a daily ﬂow threshold of 560 per

stop informs us that the losses will be contained in the

interval [3%,20%]. From ﬁgure 5 we can see that the

value of 560 corresponds to the start of the ﬁrst stage

in achieving a signiﬁcant saving in distance travelled

of 3km per day and a time saving of around 20%. This

time gain can be an opportunity to increase the buses

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

374

Figure 6: Number of lost users as function of the number

of removed stops for different walking distance threshold

services frequency. Furthermore, the additional walk-

ing distance that users will have to cover to reach a

replacement stop is between 500m and 1km.

Given that the average bus consumes between 50

and 60 litres of fuel per 100km, we can calculate a

saving of between 240L and 288L per day and for this

line. On the other hand, the CO

emission rate is in

the range of [900, 1000] g/km, thus for a reduction of

3km per trip on a line corresponds to a reduction of

emissions in the [432, 480] range kg per day for

a given bus line. If we extent the result to the bus

network of more of than 100 bus lines, this represents

several dozens of tons of CO

per day and then several

dozens of thousands of tons of CO

per year. How-

ever when bus stops are removed, users should take

another transportation mode for example bike shar-

ing, transport for demand, taxis or private car. On

the one hand, environmental parameters may favour

the use of certain transportation mode for example

the weather, congestion network, distance to cover or

proximity with city centre, etc. On the other hand,

socio-economical parameters age, household compo-

sition, employment, etc. The use of carbon modes

generates emissions that need to be taken into ac-

count. Some studies prove that people are more likely

to walk in their neighbourhood if there is a bus station

is located at less than 400m (Boulange et al., 2017).

Moreover, public transport user are more likely to

walk, between 400m and 800m, to reach their stop

(Reck et al., 2022).

The analysis of the ﬁgure 4 provides relevant in-

formations but lacks realism, as it seems that the

walking distance is the main factor. However, in our

study the walking distance threshold does not depend

on few parameters that can inﬂuence the evaluation of

the acceptability threshold. For example, the weather,

the topography of the site (slope, layout of pedes-

trian zones) or natural barriers such as a river. Hus-

sain in (Hussain et al., 2021) has already raised this

point Moreover, our assessment of the ecological im-

pact takes into account the number of people at the

stops, but the comparison in terms of CO

emissions

between car and bus takes into account a journey, i.e.

an inter-stop journey. A possible improvement to our

study would be not to take into account only the ﬂow

of people at the stops, but also to include the occu-

pancy of the bus. If occupancy is below the threshold,

we consider that the network is not optimised in terms

of CO

emissions. The model that we propose is suf-

ﬁciently simple to be implemented but there is some

phenomena that we do not consider. Firstly remov-

ing stops will increase the demand on the other bus

stops so the boarding duration will increase and so on

the pollution due to the longer dwell time. Secondly

there is some factors that can raise greenhouse gas

emissions that our model don’t take into account for

example the trafﬁc congestion, vehicle types, travel

speed or acceleration deceleration rates. Moreover

the CO

emission is calculate with (INFRAS, 2023)

which take into account vehicle type, years and type

of engine. It will be interesting to take into account

some other factors to estimate more precisely the CO

emission, for example the speed, driving style, dis-

tance or vehicle load. Thirdly, we need to know the

passenger’s behaviour, if a stop is removed, some pas-

senger will not go to the nearest stop but may be they

will use different transportation mode such as private

cars, taxis, transport-for-demand that can increase the

greenhouse gas emission.

5 CONCLUSIONS

With the world’s population constantly on the in-

crease, there is a growing demand for mobility. With

public transport systems under increasing strain, it is

important to ﬁnd ways of meeting demand efﬁciently.

At the same time, the ecological impact of transport

must also be taken into account. To this end, we pro-

pose a method for optimising a bus route in terms

of distance travelled, while measuring the impact on

quality of service and greenhouse gas emissions. Our

results show that for a ﬂow threshold value of 560

users per stop, the number of stops removed corre-

sponds to half of the stops on the line. It was found

that the rate of people who would no longer have ac-

cess to the public transport service would be between

3% and 20% depending on the accepted walking dis-

tance of 1200m and 100m respectively. With regard to

greenhouse gas emissions, the reduction in emissions

is estimated more than 450 kg per day for a given bus

Bus Routing Optimisation: A Case Study for the Toulouse Metropolitan Area

375

line which represents a reduction of more than 150

tons per line and per year. Compare to references in

the literature (Stewart and El-Geneidy, 2016) explain

that bus stop consolidation is effective and simple to

the quality of service, and conclude that remove aver-

age 23% bus stops and reducing the system coverage

area by 1%. In addition (Kehoe, 2004) show that bus

stop consolidation improve timetable reliability and

has no negative impact on ridership, and can even im-

prove it.

This study is an initial effort to examine the feasi-

bility of solutions to reduce the carbon impact on the

environment, while also assessing the impact on qual-

ity of service. Our model is not perfect and needs to

be modiﬁed to include more factors in the quantiﬁca-

tion of greenhouse gas emissions. However, we also

need to assess the inﬂuences and trends on the choice

of mode of transport when there is a modal shift.

ACKNOWLEDGEMENTS

All the mentioned authors performed the contribu-

tions to the present paper. This work was carried

out within the framework of a scientiﬁc collabora-

tion agreement with Tiss

eo Collectivit

es. The authors

thank Tiss

eo Collectivit

es for having sharing the used

data and their support.

REFERENCES

ADEME (2023). Share of public transport in total passenger

trafﬁc. https://www.odyssee-mure.eu/publications/

efﬁciency-by-sector/transport/public-transport.html.

Consulted on 2024-01-08.

Alsger, A., Assemi, B., Mesbah, M., and Ferreira, L.

(2016). Validating and improving public transport ori-

gin–destination estimation algorithm using smart card

fare data. Transportation Research Part C: Emerging

Technologies, 68:490–506.

Boulange, C., Gunn, L., Giles-Corti, B., Mavoa, S., Pet-

tit, C., and Badland, H. (2017). Examining associ-

ations between urban design attributes and transport

mode choice for walking, cycling, public transport

and private motor vehicle trips. Journal of transport

& health, 6:155–166.

Cui, A. (2006). Bus passenger origin-destination matrix

estimation using automated data collection systems.

PhD Thesis, Massachusetts Institute of Technology.

Daraio, C., Diana, M., Di Costa, F., Leporelli, C., Mat-

teucci, G., and Nastasi, A. (2016). Efﬁciency and ef-

fectiveness in the urban public transport sector: A crit-

ical review with directions for future research. Euro-

pean Journal of Operational Research, 248(1):1–20.

Dickens, M. (2023). Public transportation fact book.

https://www.apta.com/wp-content/uploads/APTA-

2022-Public-Transportation-Fact-Book.pdf. Con-

sulted on 2024-01-08.

El Mahrsi, M., C

ome, E., Baro, J., and Oukhellou, L.

(2014). Understanding Passenger Patterns in Pub-

lic Transit Through Smart Card and Socioeconomic

Data: A case study in Rennes, France.

Feng, J., Li, Y., Zhang, C., Sun, F., Meng, F., Guo, A., and

Jin, D. (2018). DeepMove: Predicting Human Mo-

bility with Attentional Recurrent Networks. In Pro-

ceedings of the 2018 World Wide Web Conference,

WWW ’18, pages 1459–1468, Republic and Can-

ton of Geneva, CHE. International World Wide Web

Conferences Steering Committee. event-place: Lyon,

France.

Ge, Q. and Fukuda, D. (2016). Updating origin–destination

matrices with aggregated data of GPS traces. Trans-

portation Research Part C: Emerging Technologies,

69:291–312.

Hora, J., Dias, T. G., Camanho, A., and Sobral, T. (2017).

Estimation of Origin-Destination matrices under Au-

tomatic Fare Collection: the case study of Porto trans-

portation system. Transportation Research Procedia,

27:664–671.

Hrelja, R., Khan, J., and Pettersson, F. (2020). How to cre-

ate efﬁcient public transport systems? a systematic re-

view of critical problems and approaches for address-

ing the problems. Transport Policy, 98:186–196.

Huang, D., Yu, J., Shen, S., Li, Z., Zhao, L., and Gong,

C. (2020). A Method for Bus OD Matrix Estimation

Using Multisource Data. Journal of Advanced Trans-

portation, 2020:5740521.

Hussain, E., Bhaskar, A., and Chung, E. (2021). Transit od

matrix estimation using smartcard data: Recent de-

velopments and future research challenges. Trans-

portation Research Part C: Emerging Technologies,

125:103044.

INFRAS (2023). Hbefa. https://www.hbefa.net/en/

software#online-version. Consulted on 2024-01-08.

Ji, Y., Cao, Y., Liu, Y., Guo, W., and Gao, L. (2019).

Research on classiﬁcation and inﬂuencing factors of

metro commuting patterns by combining smart card

data and household travel survey data. IET Intelligent

Transport Systems, 13(10):1525–1532.

Kehoe, O. d. (2004). Effects of bus stop consolidation on

transit speed and reliability: A test case. Master of

Science), University of Washington.

Li, W. (2016). Modelling bike-sharing choice in a develop-

ing country with a focus on the impacts of air pollu-

tion and weather conditions. In General Proceedings

of the 14th World Conference on Transport Research,

volume 14. World Conference on Transport Research

Society (WCTRS).

MinisterForEcology (2019). Se d

eplacer en

voiture. https://www.statistiques.developpement-

durable.gouv.fr/se-deplacer-en-voiture-seul-

plusieurs-ou-en-covoiturage-0. Consulted on

2024-01-08.

VEHITS 2024 - 10th International Conference on Vehicle Technology and Intelligent Transport Systems

376

MinisterForEcology (2022). Chiffre clef des transports.

https://www.statistiques.developpement-durable.

gouv.fr/chiffres-cles-des-transports-edition-2022.

Consulted on 2024-01-08.

Osorio, C. and Punzo, V. (2019). Efﬁcient calibration

of microscopic car-following models for large-scale

stochastic network simulators. Transportation Re-

search Part B: Methodological, 119:156–173.

Reck, D. J., Martin, H., and Axhausen, K. W. (2022). Mode

choice, substitution patterns and environmental im-

pacts of shared and personal micro-mobility. Trans-

portation Research Part D: Transport and Environ-

ment, 102:103134.

Stewart, C. and El-Geneidy, A. (2016). Don’t stop just yet!

a simple, effective, and socially responsible approach

to bus-stop consolidation. Public Transport, 8:1–23.

Sudhir Gota, C. H. (2023). Asian transport outlook.

https://asiantransportoutlook.com/documents/11/

Asian Transport 2030 Outlook Nov 2022.pdf.

Consulted on 2024-01-08.

Tiss

eo (2023). Plan tiss

eo. https://www.tisseo.fr/se-

deplacer/plans. Consulted on 2024-01-08.

UTP (2022). Observatoire de la mobilit

e 2022.

https://www.utp.fr/system/ﬁles/2022-11/

20220927 dp observatoire de la mobilite v2.0.pdf.

Consulted on 2024-01-08.

Wang, W., Attanucci, J. P., and Wilson, N. H. M. (2011).

Bus Passenger Origin-Destination Estimation and Re-

lated Analyses Using Automated Data Collection Sys-

tems. Journal of Public Transportation, 14(4):131–

150.

Bus Routing Optimisation: A Case Study for the Toulouse Metropolitan Area

377