![](bg2.png)
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