Taxi Service Simulation: A Case Study in the City of Santa Maria
with Regard to Demand and Drivers Income
Andre Brizzi and Marcia Pasin
Programa de P
´
os-Graduac¸
˜
ao em Ci
ˆ
encia da Computac¸
˜
ao, Universidade Federal de Santa Maria, Brazil
Keywords:
Simulation, Taxi Service, Drivers Income, Information System.
Abstract:
Taxi is an already well-established service in many cities around the world. Nowadays, the service request is
mainly made through mobile applications, where the user selects the desired options, including the payment
method. An information system, aware of the location of taxis, associates the closest taxi to the customer
request. In general, taxi allows the user to enjoy the mobility service without being directly charged by the
vehicle maintenance. The vehicle owner, who can be a company or a self-employed person (frequently the
taxi driver), is the one in charge with vehicle maintenance, fuel payment, etc. However, recently the taxi
service has lost much of its appeal due to competing car sharing services. Thus, it is necessary to evaluate
more carefully the implementation and maintenance of taxi service in cities with regard to the drivers income.
This work contributes in this way. Here, the taxi service is considered, with a standardized vehicle fleet but
using different vehicle types (electric, ethanol, gasoline and CNG). Given a demand and costs, a simulation is
proposed to detail and evaluate the appropriate balance between drivers income and demand scheme to keep the
service viable to the drivers. Simulation was performed in a real scenario, the city of Santa Maria in Southern
Brazil. Input values in the simulation scenario (fuel price, demand, etc.) were chosen, based on literature,
city hall documentation and Internet news, to make the simulation as realistic as possible. Simulation results
shown that for feasible taxi service, the city town hall must define a maximum number of taxi licenses. The
vehicle type has a large impact in the taxi driver’s profit. Electric vehicles have a lower cost per km driven,
but still have high cost of acquisition. Finally, if the daily traveled distance increases, the difference between
electric vehicles and the others decreases, making it possible electric vehicles to become more advantageous.
1 INTRODUCTION
Public transportation is known to be of poor quality in
many cities. And it is also known that there is a pref-
erence of passengers for individualized transportation
due to practicality, reliability, comfort, and safety. In
a recent survey (CNDL and SPC Brazil, 2017), 60.1%
of respondents who own private vehicles stated that
they would stop using it, if efficient public transporta-
tion alternatives existed. In fact, the success of pro-
posals to improve urban mobility depends on mass ac-
ceptance by users (Alazzawi et al., 2018). The avail-
ability of tools and systems that bring together differ-
ent city mobility options is a key point for tracking
and understanding the city’s mobility needs.
One way to mitigate public transportation prob-
lems is vehicle sharing services, and one of the well-
established sharing services is the taxi. Nowadays, a
taxi service request is typically made through a mo-
bile application, where the costumer selects the de-
sired options, trip origin and destination, including
the payment method. In general, taxi allows the cos-
tumer to enjoy the mobility service without having to
pay the necessary amounts for the maintenance of the
car. An information system, aware of the location of
taxis, allows the taxi closest to the customer to meet
the request. The vehicle owner, who can be a com-
pany or a self-employed person (frequently the taxi
driver), is the one in charge with vehicle maintenance,
fuel payment, etc. However, recently the taxi service
has lost much of its appeal due to competing car shar-
ing services. Thus, it is necessary to evaluate more
carefully the implementation and maintenance of taxi
service in cities.
In this work, the taxi service is considered, with
a standardized vehicle fleet. Given a fleet size and
demand, a simulation is proposed to detail and evalu-
ate the pricing scheme with regard to the taxi demand
in a city. More specifically, we assess taxi drivers
income given a city demand and with regard to dif-
ferent vehicle types (electric, ethanol, gasoline and
CNG). Thus, questions we aim answer in this paper
Brizzi, A. and Pasin, M.
Taxi Service Simulation: A Case Study in the City of Santa Maria with Regard to Demand and Drivers Income.
DOI: 10.5220/0010406800310038
In Proceedings of the 23rd International Conference on Enterprise Information Systems (ICEIS 2021) - Volume 1, pages 31-38
ISBN: 978-989-758-509-8
Copyright
c
2021 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
31
include: How many runs does the driver have to make
per day/month to be worth it? Considering differ-
ent energy sources to the vehicle engines, what is the
source of energy most profitable?
To run the simulation, we use SUMO (Behrisch
et al., 2011), a transportation network simulator with
open implementation. Simulation of the service is
performed in a real scenario, the city of Santa Maria
in Southern Brazil. In the simulation, we assume that
an information system manages the entire fleet ser-
vice and associates customers with taxis, according
to a given demand. Input values in the simulation
scenario (fleet size, demand, etc.) were chosen based
on literature, city hall documentation (Prefeitura de
Santa Maria, 2014) and Internet news, to make the
simulation as realistic as possible.
This paper is structured as following. Related
works are described in section 2. Simulation details
are described in section 3, and experiments in section
4. Conclusions are presented in section 5.
2 RELATED WORKS
In the literature, there are recent works that describe,
from the point of view of computer science and sim-
ulation, the behavior and the impact of shared vehi-
cles and taxis in transportation networks. The im-
pact of shared vehicles in the city of Milan, Italy,
was simulated with the aim of optimizing traffic by
reducing the number of vehicles circulating in streets
(Alazzawi et al., 2018). The simulation combined au-
tonomous robot-taxis, with on-demand mobility ser-
vices. Data used in the simulation include the num-
ber of vehicles circulation on the streets and mobile
cellular network usage, to model the concentration
of passengers in some areas. The simulation takes
into account the following parameters: travel time,
travel speed, waiting time for passengers to board the
robot taxi, emission of pollutants and taxi configura-
tions (with different amounts of seats). An algorithm
matches robot-taxis and consumers. According to the
authors, to eliminate congestion in Milan, it would be
necessary to reduce by 30% the number of vehicles on
the roads. To reduce demand at peak times, a dynamic
pricing system, combined with other initiatives, could
be used to motivate users to travel other time periods.
According to the seats in each car, the more seats the
robot-taxi has, the longer the costumers will have to
wait and travel due to route deviations. Robot-taxis
with around 20 seats are indicated for long distance
travel. Robot-taxis with two seats allow better travel
flexibility, but do not provide such a significant reduc-
tion in city traffic.
The combination of independent agent model sim-
ulators was also explored (Segui-Gasco et al., 2019).
MATSim (Horni et al., 2016) generates transportation
demand, associating costumers to mobility options
according to their preferences and IMSim
1
provides
an operational execution environment for transporta-
tion networks. By this combination, authors evaluated
the impact of mobility scenarios from different per-
spectives: costumers, service-operators and city hall.
The simulation was calibrated with data from London
traffic control and MERGE Greenwich Consortium
(2017-2018). Evaluated metrics were optimum vehi-
cle fleet size, vehicle type (traditional taxis and ride-
share vehicles), vehicle size (4 and 8 seating places),
vehicle occupancy, as well as wait and detour times
for each costumer. A main feature of the proposal
was the evaluation of the trade-off between quality of
service and demand. Thus, a service-operator may in-
vestigate how fleet size and energy (or even the travel
duration) affect a pricing model.
Simulation was also carried out in order to com-
pare business models for vehicle rental services (Per-
boli et al., 2017). The comparative analysis highlights
aspects of different business models and solutions ap-
plied to improve service. Business models for vehi-
cle rental services can be vehicle delivery-receipt or
free-floating. In the delivery-receipt model, fleet does
not need to be managed and relocated, but consumers
need to travel to a particular pick-up and release loca-
tion. In the free-floating model, vehicles can be re-
leased anywhere. The free-floating model tends to
better satisfy consumers, since there is no need to
travel to a particular pick-up location. However, it
requires fleet management to guarantee the availabil-
ity of vehicles in some locations, i.e., the company
needs to take vehicles that are in points of less inter-
est to places of higher demand. In this scope, different
costumers profiles can be defined: commuters (those
that travel from home to work), professional and ca-
sual. These profiles are randomly assigned to routes.
In addition, different vehicle types can be used, such
as electric and combustion vehicles. With regard to
the fleet management, electric vehicles need more ef-
fort when compared to combustion vehicles, due to
recharging time and the need to find a charging point.
Efficient route optimization was proposed as an
opportunity to increase drivers revenues (Li et al.,
2017). A vacant taxi represents wasting of both fuel
and taxi driver time. Moreover, inefficient routing
can create more traffic in the city and consequently
more pollutant are emitted. Therefore, the Markov
Decision Processes can be used to maximize drivers
revenues by the application of an efficient routing ap-
1
http://www.talon.world
ICEIS 2021 - 23rd International Conference on Enterprise Information Systems
32
Table 1: Summary of related works.
Authors Vehicle type City Simulation platform
Alazzawi et al. 2018 Autonomous and Conventional Milan SUMO/TraCI
Segui-Gasco et al. 2019 Autonomous London MATSim/IMSim
Perboli et al. 2017 Conventional Turin none
Li et al. 2017 Conventional New York none
This work Conventional Santa Maria SUMO/TraCI
inf. system
costumer
1
2
Figure 1: An information system manages taxi fleet and an-
swers client requests.
proach. Data from the taxi service from New York
City was used in the experiments. Simulation results
shown that the proposed model can collaborate to im-
prove drivers income since it reduces the time a cos-
tumer needs to find a vacant taxi.
Related works are summarized in Tab. 1. Unlike
Alazzawi et al. and Segui-Gasco et al., which simu-
late the impact of using shared vehicles in cities, seek-
ing to reduce the number of vehicles in circulation
here, such as Perboli et al., we are focusing on the
provider side. In particular, in this work we are focus-
ing on the drivers income. Unlike Perboli et al. and
Li et al., and as in Alazzawi et al. and Segui-Gasco
et al., we use simulation to investigate how different
parameters impact the expected results and drivers in-
come.
3 PROPOSED SIMULATION
In this work, taxi service is considered in a simulation
to detail and evaluate the drivers income in the end of
journeys. Fig. 1 depicts the required Information Sys-
tem (IS) to support this service. Taxis publish their
locations in the IS (1) and customers make requests
(2). The IS allocates taxis according to the location of
the customers.
The simulation scheme, implemented in SUMO
traffic simulator (Behrisch et al., 2011), is depicted in
Fig. 2 and consists of three main parts: scenario, input
Figure 2: Simulation scheme.
parameters and results. The scenario presents the map
of the geographic region to be simulated. The map has
tow layers: a static layer and a dynamic layer. The
static layer is previously obtained through a cut in the
map of Open Street Map
2
, exported in .osm format.
Using the SUMO Simulator script, the osm file
is converted into a transportation network, a scenario
formatted to be simulated by SUMO. The network is
composed by edges (street corners) and connections
between edges (street blocks). In the scenario conver-
sion, the path to the .osm file is indicated and addi-
tional parameters, such as the generation of sidewalks
along the roads can also be informed. After complet-
ing the stage of generating the map scenario, the out-
put is a file in .net.xml format (i.e., the description of
the transportation network). This map runs in a server
in which other simulation parameters can be config-
ured. For instance, the duration of the simulation.
The simulation parameters are sent to the simula-
tion server via Traffic Control Interface (TraCI) (We-
gener et al., 2008). On the server side, the parameters
are used by the simulator generate the demand (i.e.,
the taxi runs) that associate costumers and taxis. Us-
ing the randoTrips.py script, provided by SUMO, ran-
dom trips are automatically generated, both for cos-
tumers and vehicles. We have the possibility to define
2
https://www.openstreetmap.org
Taxi Service Simulation: A Case Study in the City of Santa Maria with Regard to Demand and Drivers Income
33
parameters for this script such as:
maximum distance that a costumers can walk,
probability that a trip can start at the scene, and
vehicle intensity flow and costumers/pedestrians
flow and, in addition, to establish which vehicle a
costumer can choose to complete her/his journey
trip.
The randoTrips.py script generates a file in the
.rou.xml format with valid routes to be used by
SUMO. The next step is running the simulation. The
SUMO simulations are presented by a .config file,
which contains the name of the file with the sce-
nario, .net.xml, of the additional items, .add.xml and
of the routes, .rou.xml. When loading the simulation,
SUMO searches for the information in the files pro-
vided. Also in the .config file, it is possible to define
the output to be presented after the simulation.
With regard to our simulation, some output in-
formation can be obtained automatically by SUMO
and include, for example, the vehicle average speed.
However, some specific routines have been coded,
since SUMO does not implement all the necessary
routines required in the scope of this work.
In general, simulation results that we are mainly
investigated in our scenario include:
gross and net drivers incomes, with regard to the
number of runs, and
drivers incomes, with regard to the vehicle type
(engine).
To summarize, the developed simulation receives
the data from the simulation files, and presents the
resulting values that we discuss in more details in the
section 4.
Parameters in our simulation defined according to
real-world available data to drivers/taxis include:
price of the fuel, vehicle model, etc.,
monthly rental amount and vehicle consumption
related to the taxi model,
formula to compute the payment for a taxi run,
which is composed by a fixed amount and the
amount per km traveled,
working hours for drivers, and
number of available taxis.
Reference values are shown in Tab. 2. These val-
ues directly influence the driver’s revenue. Parameters
that can be defined in the simulation, using city/traffic
information, according to real-world observations in-
clude:
intensity of the vehicles flow in the scenario to be
defined by counting the number of vehicles in a
given simulation interval,
average travelled distance, defined according to
the behavior of costumers in that region, and
demand for runs, which can be calibrated using
information provided by city hall.
At the end of the simulation, the travel cost can
be computed and, therefore, the drivers income. The
travel cost depends on the period of the day and the
distance driven by the taxi driver.
4 EXPERIMENTS
In this section, we first present the scenario setup tak-
ing into account the city of Santa Maria, then we de-
scribe the demand generation process, i.e., the addi-
tion of costumers in the simulation interested in rid-
ing a taxi. In the following, we describe the simula-
tion process, and finally, we focus on the simulation
results of our experiments.
4.1 Scenario Setup
In Santa Maria downtown, there are 14 taxi stops
which are part of our simulation map. We assume
that half of these points have 2 taxis and the other
half have 3 taxis, resulting in a total of 35 taxis in the
simulation. Fig. 3 presents a screenshot of our simu-
lation environment in SUMO, with a set of streets in
the center of Santa Maria city.
In general, the city has a very irregular layout in
its streets. Each red diamond in the green map rep-
resents a taxi station. Each blue square represents a
pick up or an unboarding point, manually chosen for
this simulation.
4.2 Demand Generation
The pedestrian (costumer) demand in our simulation
is generated by PersonFlows routine from SUMO, in
which people are inserted at different points on the
map. This component periodically generates pedes-
trians in a defined location. Pedestrians follow a pre-
defined route to reach their destination, being able to
get around on foot or using a taxi vehicle. Other vehi-
cles are not inserted in the simulation, but the effect on
the traffic behaviour of the other vehicles (bus, trucks
and private vehicles) is due to the configured speed
limitation that the taxi can develop in the city.
The simulation in SUMO takes place so that the
vehicles present in a routing list are inserted in the
ICEIS 2021 - 23rd International Conference on Enterprise Information Systems
34
Figure 3: Screenshot of SUMO simulation environment.
simulation at the given time and after completing the
route these vehicles are removed from the simulation.
In order to make it possible for a taxi to perform mul-
tiple trips during the simulation, it is necessary to use
an auxiliary script to generate new routes for this ve-
hicle during the simulation. The script used for this
purpose is the Demand Keeper. This script is part of
the Net Populate
3
project, a set of scripts for gener-
ating and controlling demand in SUMO experiments.
Finally, Demand Keeper call is used in conjunction
with the TraCI interface, which allows to interact in
real time with SUMO.
4.3 Simulation Parameters and Drivers
Income
In the simulation, taxis are set to be in service from
8:00 a.m. to 4:00 p.m. Thus, we consider each taxi
service operates with three driver shifts per day, and
each taxi journey has the duration of 8 hours. Each
step of the simulation represents one second of time,
so the total number of steps in the simulation is 28,800
(i.e., 8 simulation hours). We run the simulation four
times, each time with an average demand for taxi rides
(i.e., 5, 10, 20 and 30 rides/day in average per taxi).
These values for the number of runs were chosen for
only 5 runs per day per driver to reflect a lockdown
scenario due to the new coronavirus pandemic, for in-
stance, and 30 runs would be a more optimistic sce-
nario. The number of costumers in each simulation
is modeled in order to create an average number of
rides per taxi in each simulation run. For simulation
purpose, we consider a standardized vehicle fleet.
Simulation parameters are summarized in Tab. 2.
Each taxi run starts with an initial value called flag
3
https://github.com/maslab-ufrgs/net-populate
B
i
, in which i = {0, 1, 2}, given the day of the week
and time, and the cost per kilometer traveled. The
values charged by the taxi drivers are stipulated by the
city hall (Araujo, 2020). Driver expenses also include
the fuel consumption per litre C, the maintenance cost
per kilometer traveled M, insurance expenses I, and
vehicle loan P.
Table 2: Simulation parameters.
Symb. Parameter Value
B
0
flag-down fare 5.64 BRL/km
B
1
flag-down fare 1 3.36 BRL/km
B
2
flag-down fare 2 4.03 BRL/km
G fuel price 4.50 BRL/L
C fuel consumption tax 10 km/L
M vehicle maintenance 0.20 BRL/km
I annual insurance 2,000.00 BRL
P vehicle loan (monthly) 600.00 BRL
In addition to the parameters of Tab. 2, we add
that taxis move at an average speed of 36 km/h. For
the costumer-taxi association, we use the algorithm
implemented by SUMO where the taxi closest to the
costumer wins the run. In the simulation, we calculate
the gross income average obtained by taxis drivers
during the 8 hours of work, using Eq. 1 to compute
the individual Gross Income (GI) for each taxi driver:
GI = R · B
i
+ D
t
· B
i
, (1)
where R means taxi runs for the driver and D
t
means the total of the traveled distance (km). We also
compute the Net Income (NI) per taxi driver, using
Eq. 2, which is obtained by subtracting the vehicle
expenses from the gross amount, given by:
NI = GI D
t
·
G
C
P
30
I
365
. (2)
Taxi Service Simulation: A Case Study in the City of Santa Maria with Regard to Demand and Drivers Income
35
0
50
100
150
200
250
300
350
5 10 20 30
Income (BRL)
Runs per working day
GI
NI
Figure 4: Values for GI and NI obtained in the experiment
using the parameters described in the Table 2.
We do not consider the amount paid in mainte-
nance of the vehicle in our equations, but this value
should be considered in a future study. In fact, some
values such as maintenance, insurance and financing
can be shared by drivers who drive the same vehicle.
In the following, we highlight the simulation re-
sults of computing net and gross incoming and for taxi
drivers. We evaluate two different aspects: simulation
results with regard to drivers income and simulation
results with regard to the vehicle energy source.
4.4 Simulation Results with Regard to
Drivers Income
Here we assess the drivers income in different sce-
narios, from the pessimistic to the more optimistic.
Tab. 3 shows the simulation results for the average
travelled distance for costumers and drivers, given
different amounts of taxi runs.
Table 3: Travelled distance with regard to costumers and
total average distance, per driver.
Taxi Costumers Drivers total
runs (R) dist. avg. (km) dist. avg. (km)
5 7.7 12.6
10 14.9 21.5
20 29.8 45.5
30 45.2 65.8
In our simulation, the average distance traveled in
each trip is 1.5 km. In the most pessimistic scenario
(5 runs), the driver drives only 12.6 km per day, and
in the most optimistic scenario, the driver drives 65.8
km per day. Considering both scenarios (pessimistic
and optimistic), Fig. 4 depicts the (average) gross and
net values (GI and NI) obtained by the drivers per day
depending on the number of runs performed.
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
4 6 8 10
Income by month (BRL)
Working hours per day
5 daily runs
10 daily runs
20 daily runs
30 daily runs
Figure 5: Monthly NI based on working hours.
In Fig. 5, income values are plotted by month,
with regard to different number of working hours.
We, in particular, extrapolate the depicted values to
10 working hours. It is clear that the more the driver
works, the more she/he earns. However, if demand
is not enough, the driver is unable to pay the service
costs.
From the simulation results depicted in Figs. 4 and
5, we may conclude that it is impracticable to provide
taxi service in scenarios where the demand for taxi
runs is only 5 daily. Only 5 runs results in a monthly
gain of 952.24 BRL, less than the minimum wage cur-
rently in force by Brazilian legislation, given the law
number 14,013 (Brazil, 2020), which is 1045.00 BRL.
In contrast, if the taxi driver works in periods with de-
mands of 10 daily taxi runs, it is possible to guarantee
to the taxi driver an income above the minimum wage
working only 4 hours a day.
Given these results, it is important to highlight the
importance of balancing the amount of taxi licenses
allowed by the city hall and demand, in order to guar-
antee a sufficient number of vehicles to serve passen-
gers while allowing the activity to remain profitable
for taxi drivers. It is also important to observe that in
pessimistic scenarios, such as lockdown scenarios, for
instance, government contributions need to be consid-
ered for taxi service providers.
Another important observation is about the devi-
ation pattern we computed to the drivers income. In
our experiments, the deviation was quite high, as the
routing algorithm always ends up choosing the same
taxis that are closest to the passengers while other
taxis barely manage to run. For scenarios with high
demand, there is a greater turnover between taxis and
traveler origin points and destination points. There-
fore, a new algorithm to associate taxis and pedestri-
ans needs to be proposed in future work.
ICEIS 2021 - 23rd International Conference on Enterprise Information Systems
36
4.5 Simulation Results with Regard to
the Vehicle Energy Source
In addition to demand, another factor that impact the
taxi drivers income is the vehicle energy source. Here,
we consider three types of vehicles, according to the
energy sources:
flexible-fuel (flex) vehicles, which are capable of
running with gasoline and ethanol,
bi-fuel vehicles, which engines are capable of run-
ning on two fuels: a internal combustion engine
(with gasoline or diesel), and the other alternate
fuel such as natural gas (CNG), and
electric vehicles, which are charged through the
electric power network.
In order to keep the vehicle in good condition, the
taxi driver changes her/his vehicle for a new one ev-
ery 5 years. Assuming that the taxi driver has a flex
vehicle that is completing 5 years of use and needs
to change for a new one, she/he can choose from the
three types of vehicles mentioned above.
To purchase a new vehicle, we consider that the
driver current vehicle is worth 30,000.00 BRL, which
is used as an input for financing. The financing rate
is 1% monthly on average and the financing term is
60 months. We emphasize that in Brazil, new vehi-
cles purchased by taxi drivers have tax incentives that
resulting in a value up to 30% less than paid by an or-
dinary consumer. We also considering that, in case of
CNG as energy source, typically, a conversion kit is
installed in the taxi and allows an originally flex vehi-
cle to be supplied with CNG. The cost of installing a
CNG kit in a vehicle is 5,000.00 BRL on average.
To allow the evaluation of energy source, we con-
sider other values described in Tabs. 4 and 5. Tab. 4
shows the different types of vehicles and the respec-
tive installment to be paid. For flex vehicles, the price
of the Renault Logan was considered, presented by
the manufacturer’s website in October 2020, for sale
with exemption for taxi drivers. The electric vehi-
cle chosen to the simulation was the one with the
lowest value found for sale currently in Brazil, JAC
iEV20. The price we used was according to the man-
ufacturer’s website in October 2020, considering an
exemption of 30% of the value for taxi drivers.
The choice of vehicle type in order to maxi-
mize driver profit must take into account acquisition
cost and the cost per kilometre for travel. Tab. 5
presents vehicles comparison cost per travel kilome-
ter. Values for electric vehicle consumption we con-
sider here are based on the literature (Besselink et al.,
2011). Fuel price here used is based on the price
national survey carried out by the Brazilian National
0
50
100
150
200
250
300
350
5 10 20 30
Net income (BRL)
Rides per working day
electric engine
ethanol
gasoline
CNG
Figure 6: Drivers NI with regard to energy source.
Petroleum Agency
4
, relative to October 2020. In gen-
eral, flexible-fuel vehicles have higher maintenance
costs when compared to electric vehicles (Alexander
and Davis, 2013). In contrast, electric vehicles have a
considerably high acquisition cost when compared to
vehicles with internal combustion engines.
Fig. 6 depicts our simulation results to drivers NI
based on energy source. Among the energy sources
we analyzed, it is possible to state that CNG max-
imizes the NI of taxi drivers in the scenarios of 20
and 30 daily taxi runs and ties with gasoline in the
scenario of 10 runs. For 5 daily runs, gasoline pro-
vides the highest profit, with CNG being affected in
this scenario by the cost of installing the conversion
kit, which reflects in a higher installment value. Al-
though ethanol has a lower cost per liter than gasoline,
its autonomy has resulted in a lower NI than gasoline
in all scenarios.
Actually, the use of electric vehicles is not profit
to taxi drivers when there are only 5 daily runs, but
as the number of daily runs increases, the difference
with regard to the profit in relation to other energies
decreases. With 30 daily runs, the electric vehicle has
a profit similar to a vehicle with ethanol. Although
it has the lowest cost per km traveled among all the
considered energies, the electric vehicle still has high
acquisition cost that results in large fixed expenses,
harming the taxi driver’s NI.
5 CONCLUSIONS
This work proposed the evaluation of the taxi service
from the point of the view of the taxi driver (income).
The evaluation was conducted with simulation sup-
port, and considering an information system to deal
with entire taxi fleet service and to associating cus-
tomers with taxis. From a real scenario, a simulation
4
http://preco.anp.gov.br
Taxi Service Simulation: A Case Study in the City of Santa Maria with Regard to Demand and Drivers Income
37
Table 4: Estimated vehicle acquisition cost.
Vehicle
type
First installment
(BRL)
Rate % a.m. Term
(months)
Final price (BRL) Installment
(BRL)
Electric 30,000.00 1 60 98,000.00 1,500.00
Flex 30,000.00 1 60 42,000.00 267.00
CNG 30,000.00 1 60 47,000.00 378.00
Table 5: Fuel comparison costs, given in kilometre per litre (Total) with regard to vehicle maintenance.
Vehicle type Price Mileage Maint. (BRL/km) Total (BRL/km)
Electric 0.50 (BRL/kWh) 0.2 (kWh/km) 0.10 0.20
Flex (ethanol) 4.00 (BRL/L) 7.0 (km/L) 0.20 0.77
Flex (gasoline) 4.50 (BRL/L) 10.0 (km/L) 0.20 0.65
CNG 3.72 (BRL/m
3
) 12.3 (km/m
3
) 0.20 0.50
from a service journey was performed.
Simulation results shown that for the taxi service
to be feasible for drivers, the city town hall must
define a maximum number of taxi licenses in order
to ensure that the average daily travel per driver is
not less than 10 runs. Smaller values mean that the
monthly taxi gain does not reach the minimum wage
established by the Brazilian government. The mini-
mum number of taxis allowed in a region must con-
sider the quality of the service, so as not to compro-
mise the availability of the service to costumers.
Vehicle type has a large impact in the taxi driver’s
profit. Simulation results showed that although elec-
tric vehicles have a lower cost per km driven, the high
cost of acquisition made the taxi driver’s net profit re-
sult in lower values than other types of vehicle. Fi-
nally, we mention that as the daily traveled distance
increases, the difference between electric vehicles and
the others decreases, making the new technology to
become more advantageous.
REFERENCES
Alazzawi, S., Hummel, M., Kordt, P., Sickenberger, T.,
Wieseotte, C., and Wohak, O. (2018). Simulating the
impact of shared, autonomous vehicles on urban mo-
bility - a case study of Milan. SUMO 2018- Simulat-
ing Autonomous and Intermodal Transport Systems -
EPiC Series in Engineering, 2:94–110.
Alexander, M. and Davis, M. (2013). Total cost of own-
ership model for current plug-in electric vehicles. In
TECHNICAL REPORT 3002001728. Electric Power
Research Institute (EPRI).
Araujo, M. (2020). Prefeitura de Santa Maria concede, ap
´
os
tr
ˆ
es anos sem aumento, reajustes nas tarifas de t
´
axis.
https://www.santamaria.rs.gov.br/noticias/20395-
prefeitura-de-santa-maria-concede-apos-tres-anos-
sem-aumento-reajustes-nas-tarifas-de-taxis.
Behrisch, M., Bieker, L., Erdmann, J., and Krajzewicz, D.
(2011). SUMO - simulation of urban mobility (an
overview). In Proc. 3rd Int. Conf. on Advances in Sys-
tem Simulation (SIMUL 2011), pages 63–68.
Besselink, I. J., Hereijgers, J., Van Oorschot, P., and Nijmei-
jer, H. (2011). Evaluation of 20000 km driven with a
battery electric vehicle. EEVC - European Electric
Vehicle Congress, Bruxelles, pages 1–10.
Brazil (2020). Lei n
o
14.013, de 10 de junho de 2020.
CNDL and SPC Brazil (2017). H
´
abitos e Prercepc¸
˜
oes sobre
a Mobilidade Urbana no Dia a Dia dos Brasileiros.
SPC Brazil.
Horni, A., Nagel, K., and Axhausen, K. W. (2016). The
multi-agent transport simulation MATSim. Ubiquity
Press London.
Li, P., Bhulai, S., and van Essen, J. T. (2017). Optimization
of the revenue of the New York city taxi service using
markov decision processes. In Proc. 6th Int. Conf. on
Data Analytics, pages 47–52, Barcelona (Spain).
Perboli, G., Ferrero, F., Musso, S., and Vesco, A. (2017).
Business models and tariff simulation in car-sharing
services. Transportation Research Part A: Policy and
Practice, 115.
Prefeitura de Santa Maria (2014). Lei municipal n
o
5.863,
de 09 de maio de 2014.
Segui-Gasco, P., Ballis, H., Parisi, V., Kelsall, D. G., North,
R. J., and Busquets, D. (2019). Simulating a rich
ride-share mobility service using agent-based models.
Transportation 46, pages 2041–2062.
Wegener, A., Piorkowski, M., Raya, M., Hellbr
¨
uck, H., Fis-
cher, S., and Hubaux, J.-P. (2008). TraCI: An interface
for coupling road traffic and network simulators. 11th
Communications and Networking Simulation Sympo-
sium (CNS), pages 155–163.
ICEIS 2021 - 23rd International Conference on Enterprise Information Systems
38