Content uploaded by Zafeiris Kokkinogenis
Author content
All content in this area was uploaded by Zafeiris Kokkinogenis on Mar 10, 2019
Content may be subject to copyright.
Traffic Simulation of Lane-Merging of Autonomous
Vehicles in the Context of Platooning
Gil Domingues, Jo˜
ao Cabral, Jo˜
ao Mota, Pedro Pontes, Zafeiris Kokkinogenis, and Rosaldo J. F. Rossetti
Artificial Intelligence and Computer Science Lab
Department of Informatics Engineering
Faculty of Engineering, University of Porto
Rua Dr Roberto Frias, s/n — 4200–465, Porto, Portugal
{gil.domingues, up201304395, up201303462, pedro.martins.pontes, pro08017, rossetti}@fe.up.pt
Abstract—Recent research has shown that by optimizing lane
changing maneuvers and making vehicles travel much closer to
each other it is possible to achieve a significant reduction in
traffic congestion – a severe problem in all major cities. The issue
of lane-merging was approached in the context of autonomous
vehicles, capable of inter-vehicle communication and grouped
into platoons. Solutions based on the use of negotiation techniques
were explored by modelling and performing simulations in SUMO
(Simulation for Urban Mobility). This paper demonstrates the
benefits of the use of platoons in traffic flow, across a range of
metrics, and proposes two methods of negotiation for vehicles
entering a lane to merge into a platoon.
Index Terms—Autonomous Vehicles, Inter-vehicle Communica-
tion, Lane Merging, Negotiation, Simulation, Vehicle Platooning.
I. INTRODUCTION
Traffic congestion in road networks is a problem common
to all major cities, as the number of vehicles on the roads
continues to increase. This is an issue that is aggravated by
improper vehicular maneuvers such as lane changing and lane
merging.
A promising approach to tackle this problem consists in
increasing road capacity by making vehicles travel much closer
to each other using platoon-based formations. In order to
ensure the effectiveness and safety of the platooning operations
research efforts currently focus on the autonomous vehicle
technology. The removal of human factors (perception errors,
reaction time, and impulsiveness among others) allows us
to achieve the optimization of lane-merging maneuvers and
thus to improve traffic congestion [1]. Studies have shown
that platooning and lane-merging approaches could allow for
an increase in road capacity as high as 770% depending on
the number of vehicles in the formation, their velocity and
the spacing between them [1]. In addition, a close-following
formation allows for more fluid circulation, as well as up to an
8.5% saving in fuel consumption under certain conditions [2].
In this paper, we will focus on the problem of lane merging
in a situation where a vehicle intends to join an existing
platoon. We consider a scenario in which a secondary way
merges into a principal road. In this topological setup, vehicles
appear on the secondary way and attempt to join the vehicle
groups on the principal one. The operation may require the
creation of a slot in the middle of the traveling platoon. This
can lead to situations in which the merging vehicle’s intention
is in conflict with the platoon’s best interest that is not to alter
its speed significantly, so as not to worsen the arrival times
of its members too. We will consider a negotiation approach
to ensure conflict resolution and coordination among vehicles
in merging operations. A slot’s creation would be conditional
to the platoon’s performance loss being outweighed by the
value of the proposal made by the merging vehicle. There-
fore, for this approach to be feasible in a scenario involving
autonomous vehicles, inter-vehicular communication and the
establishment of a sort of economy are paramount.
Despite the recent technological advances in the field of
autonomous driving, research within the context of cooperative
autonomous vehicles is somewhat limited. Therefore, this
work stands as a potentially valuable resource within the field
of cooperation of autonomous vehicles, since it sheds some
light upon the possibility and efficiency of applying different
negotiation techniques to lane-merging situations.
The main purpose of this paper is to study cooperation
aspects in platooning/lane-merging situations assuming au-
tonomous or semi-autonomous communication-enabled vehi-
cles. A secondary purpose consists of a preliminary assessment
on the efficiency of a naive-based negotiation approach in
reducing key metrics related to traffic congestion. To this
end, two components were developed: The first is an agent-
oriented environment where the simulation’s negotiation logic
was implemented using Python. The second component is
comprised of a set of simulation scenarios to be run in SUMO,
a microscopic traffic simulator. The two components were
connected using TraCI, an API that enables interaction with
SUMO simulations at run-time.
The contributions this work intends to provide are twofold.
Firstly, we examine negotiation approaches for fair and ef-
ficient lane-merging and platoon management. We attempt to
discuss the efficiency of a naive-based (bargaining) negotiation
technique for the aforementioned scenarios. Secondly, we
aim to model basic lane-merging/platooning maneuvers in
distinct ways, so as to test and support our claims within the
research context of cooperative and communication-enabled
autonomous vehicles with platooning capabilities.
The remainder of the paper is structured as follows. First,
the literature related to platooning and lane-merging models
is summarized. Then, the methodological approach followed
is outlined. Next, the implementation is detailed. Following
that, the results obtained are presented and analyzed. Finally,
conclusions and possibilities for future work are presented.
II. RE LATE D WOR K
Within the field of traffic simulation [3], there is already a
significant amount of literature concerning both platooning and
autonomous vehicles and their interaction. Nevertheless, there
are still some interesting topics in need of further study and
analysis, namely the process of negotiation between a platoon
and a vehicle wishing to merge into it, which is the subject of
this paper.
Despite the fact that there is considerably more literature
regarding intersection management than lane merging man-
agement available, a parallel can be drawn between the two,
as lane merging can be considered as a particular type of inter-
section. Some studies focus on the use of traffic control devices
[4] to manage traffic flow on the physical traffic network.
In this paper, the authors discuss the use of reinforcement
learning in non stationary environments. Such approaches are
useful to environments where cohabitation between regular
vehicles and autonomous vehicles is required but preclude
agent-based negotiation by excluding the possibility of a
network exclusively composed of autonomous vehicles.
Even when solely considering autonomous vehicles, there
exists literature which considers either the possibility of hav-
ing a mediator – usually an agent based representation of
the intersection – responsible for managing the negotiation
between vehicle agents [5], or the direct negotiation between
vehicle agents, without any mediator agent [6]. Specifically,
the aforementioned paper suggests a game-theory approach to
right-of-way conflicts in lane-merging scenarios.
Auction-based approaches to time-space-slot allocation at
intersections have been analyzed in the context of a traffic-
based infrastructure based on autonomous vehicles [5]. Further
research work focused on ensuring fairness in a competitive
market system by considering factors such as vehicle distance
to the intersection, current speed, lane position and direction
of intended turn [7].
In the area of autonomous vehicle platooning, research has
been published [8] on scheduling platoons arriving on a k-
way merge intersection or traffic crossings in such a way as
to preserve both platoons. However this research does not
consider the possibility of vehicles joining a platoon at the
intersection.
The work by Niu et. al [9] explores the effectiveness of a
centralized coordination agent to orchestrate a complex multi-
agent system. This implementation is similar to the proposed
solution of using a single autonomous vehicle as the platoon
leader.
Platooning management techniques that have been the focus
of research include protocols based on Vehicular Ad-hoc
Networks (VANET), where vehicles establish Ad-hoc networks
through the use of wireless communication technologies.
Amoozadeh et. al [10] propose a platoon management protocol
consisting of three main primitive operations: merge, split,
and lane change. Through the composition of these three
operations other common behaviours can be implemented. The
authors propose a centralized approach to platoon coordination
through the use of a single platoon member designated as
leader to make decisions on behalf of the platoon, extolling
the privacy and performance advantages.
The idea of using auction-based negotiations between au-
tonomously driven vehicles for the purpose of determining
priority in scenarios such as overtakes and intersections has
also been explored [11]. This work differs slightly from the
the ones mentioned above, since it involves applying agent-
based negotiation to a lane-merging operation in settings of
autonomous vehicles platoon formation.
III. METHODOLOGICAL APP ROAC H
There are a few approaches through which the problem
of optimizing travel time and fuel consumption using agent-
based negotiations can be addressed. The first and most basic
approach would be to conduct a one-on-one negotiation with
a potential platoon-leader in order to reserve a slot in the
platoon. This concept can be further enhanced by having
the platoon-leader act according to the needs of each of the
elements within the platoons and not just its own.
As a starting point, a reference model will be devised,
consisting in a lane merging situation without platooning or
negotiation. In a second scenario, a vehicle attempts to merge
into a lane where platoons transit; again, no negotiation takes
place, and the platoons do not allow any vehicle to join. A
third scenario consists in a vehicle attempting to merge into a
platoon, requesting permission to do so – to which the platoon
responds with a proposal which may be either accepted or
declined (take-or-leave). Finally, a fourth scenario consists in
a vehicle attempting to merge into a platoon, negotiating the
conditions for doing so.
Controllable variables in the simulation experiments include
the number of vehicles, the road configuration, the vehicle’s
maximum, minimum and desired velocity, as well as its
maximum acceleration, the maximum number of vehicles per
platoon and minimum safety distance among them. Each
simulation run generates a variety of metrics, namely, the
average fuel consumption and CO2emissions, the average
amount of time spent stopped and the average amount of time
lost by not traveling at the desired velocity.
In scenarios in which a vehicle attempts to merge into a
platoon by either requesting permission to do so or negoti-
ating the conditions of its entrance into the said platoon, a
negotiation protocol is required. In the first case, the protocol is
straightforward: the vehicle informs the platoon of its intention
of joining the platoon. The platoon should reply with either a
refusal or the conditions to join the platoon, which the vehicle
can accept or refuse. In the second case, the protocol is some-
what more complex: the vehicle must negotiate its entrance
into the platoon by sending the platoon a request to join it, to
which the platoon leader should reply with the conditions to
joining it. The vehicle can accept those conditions or try to
renegotiate those conditions, by making a counter-proposal.
In order to simulate the negotiation mechanisms described,
an agent-based metaphor is conceived, in which each vehicle
determines its own behaviour and is capable of handling such
negotiations when required. Two entities are identified when
considering the system, specifically, the autonomous vehicle
and the platoon (Tables I and II, respectively). The platoon
represents a set of autonomous vehicles moving at a constant
speed while maintaining a reduced space between them. The
possibility of communication between any two vehicles with
a negligible transmission latency is one of the assumptions
made.
TABLE I
ATTR IBU TES D EFI NIN G TH E ENT IT Y Vehicle.
Attribute Description
Current Velocity Vehicle’s current velocity
Minimum Velocity Minimum velocity the vehicle accepts to travel at
Desired Velocity Velocity the vehicle prefers to travel at
Maximum Velocity Maximum speed the vehicle accepts to travel at
Maximum
Acceleration Maximum acceleration a vehicle tolerates
Budget Vehicle’s credit for the negotiations
TABLE II
ATTR IBU TES D EFI NIN G TH E ENT IT Y Platoon.
Attribute Description
Vehicles Platoon’s current members
Minimum Safety Distance Minimum distance to be kept between each
consecutive vehicle belonging to the platoon
Maximum Size Maximum number of members
A. Reference Model
The reference model consists of a lane-merging situation
without platooning or negotiation, where a vehicle attempts to
merge into a lane where traffic is flowing regularly.
In this basic scenario, the vehicles roam freely,
independently from each other, on the main lane of the
road stretch, following the standard rules of traffic flow. This
is the scenario which more faithfully represents how vehicles
currently behave on traffic networks.
In order to experiment with multiple scenarios and to assess
the system’s response to various vehicle behaviours, distinct
hypothetical settings – in which the different entities exhibit
different behaviours – were devised, and simulations were
conducted using the following scenarios:
B. Base Scenario
This scenario differs from the reference model by contem-
plating platooning. This way, a vehicle attempts to merge into
a lane where platoons transit. Still, no communication takes
place between the vehicle and the platoon.
C. No-Negotiation Scenario
The vehicle attempting to merge into a platoon requests
permission from the platoon, to which the platoon responds
with the position in which the car should enter it.
In the no-negotiation scenario, a take-or-leave proposal is
considered as the coordination mechanism. Here, a merging
vehicle will be accepted provided that the merging maneuver
is possible. The merging vehicle will, if accepted, join the
platoon, taking a position recommended by the platoon as
being the most convenient to do it. In this scenario, no
negotiation takes place, since the merging vehicle is subject
to the conditions imposed by the platoon, such as the position
in which it can enter the platoon or the speed at which it will
then circulate.
D. One-To-One Negotiation Scenario
The vehicle attempting to merge into a platoon proposes a
value for his permission to enter the platoon, which develops
into a negotiation between the vehicle and the platoon. Once
the platoon and the vehicle reach an agreement the vehicle is
allowed to enter the platoon or, if no agreement is reached,
the vehicle is denied entrance.
In the bargaining negotiation scenario, the vehicle makes a
proposal for its entry into the platoon, composed by its desired
position and a value which represents how useful merging into
that particular position of the platoon is to that vehicle. This
value is determined based on both the discrepancies between
the vehicles’ interests and the position in the platoon in which
the vehicle wishes to enter, and can be represented by the
following utility function:
Uv=Ec ×Budget
1 + |MaxVdif |+|M inVdif |+|CurVdif |+|P osdif |
where Ec represents the ratio between the maximum amount
of consecutive route edges the merging vehicle has in common
with a vehicle already in the platoon and the number of edges
the merging vehicle’s route contains, Budget is the credit
available for the merging vehicle to spend in its trip with
negotiations, MaxVdif is the difference between the vehi-
cle’s maximum velocity and the platoon’s maximum velocity,
MinVdif is the difference between the vehicle’s minimum
velocity and the platoon’s minimum velocity, CurVdif is
the difference between the vehicle’s current velocity and the
platoon’s current velocity, and finally P osdif is the difference
between the vehicle’s proposed entering position and the
entering position proposed by the platoon. This last variable,
P osdif is only relevant in case the platoon presents a counter-
offer with a different position, otherwise its value is considered
to be 0.
The platoon determines the value it considers appropriate for
the vehicle to pay in order to enter the platoon in the requested
position. For this purpose, a cost function is constructed to
calculate said value, by factoring in the position requested
and the resulting impact in the vehicle – in regard to velocity
variations, as well as the value of the offer made by the
merging vehicle, which is as follows:
Cp=2P Vv
(1 + P osv
Nv)0.5
where P Vvis the value proposed by the vehicle, that is the
value which is calculated using the vehicle’s utility functions.
P osvrepresents the position in which the vehicle intends to
enter the platoon and Nvis the number of vehicles already in
the platoon.
It is important to note that the platoon’s cost function
makes use of the vehicle’s proposed value, calculated with
its utility function. This offers a guarantee that difference
becomes neither disproportionately larger nor smaller.
Once the two values have been calculated and exchanged
between the two entities, a bargaining type process begins.
Each entity increments or decrements their proposed value
progressively, converging towards an agreement. Each one has
a specific negotiation step, which represents the difference
in value between two consecutive proposals by that entity.
The proposal mechanism used in each negotiation has a
corresponding Time-To-Live value (TTL), which limits the
maximum number of proposals sent for an agreement to be
reached; in case this value drops below 0 before an agreement
is reached, the vehicle will not join the platoon. This happens
when either the vehicle or the platoon is not flexible enough
– having a small negotiation step – or when the difference
between the vehicle’s measure of utility and the platoon’s
measure of cost is too great.
IV. IMPLEMENTATION
Following previous research efforts [12], [13], two compo-
nents have been developed. The first is an agent-based environ-
ment where the simulation’s negotiation logic is implemented
following an agent-oriented approach, in which the different
entities – vehicles and platoons – are implemented as separate
Python classes that emulate agent behaviour, carrying out
negotiation processes and maneuverability operations when
necessary.
The second component is comprised of a set of simulation
scenarios to be run in SUMO – a microscopic traffic simulator.
SUMO can be connected with other applications through
TraCI, a tool that allows for the issuing of commands into
SUMO. This tool is TCP-based and follows a client-server
architecture, granting our system the ability to influence the
simulation at run-time.
Fig. 1 provides a logical overview of the solution’s archi-
tecture. There are four main parts to the system, excluding
SUMO scenarios. One is the Platoon class, which allows the
enforcement of platoon-like behaviour on a set of vehicles.
This class contains directives ranging from gap initiation to
maneuver calculation and merging negotiation, which varies
depending on the scenario.
The Vehicle class represents an autonomous driving vehicle
and comprises a set of data regarding a vehicle, such as its
intended velocity and the platoon it belongs to.
Fig. 1. Logical view.
The SimStepListener class is the core of our system. This
component performs control operations at every tick of the
simulation and initiates gaps and merges when necessary.
Finally, each scenario has a launcher program – Runner –
through which the simulation is launched and the vehicles
spawn times and routes are scheduled. This component joins
the functionality of all others.
In each scenario, platoons are scheduled to spawn continu-
ously within 100 simulation ticks of each other. Individual
vehicles spawn according to a Poisson distribution with a
λ= 40 ticks. The merging window is the duration in which a
merge is possible. When the leading vehicle in a platoon steps
onto the first induction loop, that platoon is marked as being
in the merging window. As such, any vehicle stepping over
the induction loop in the merging lane will be able to request
entry into this platoon. As soon as the leading vehicle steps
on a third induction loop, at the end of the main lane, the
merging window is considered to be closed, and any vehicle
coming from the merging lane will either request entry into
a different platoon that has entered the merging window or
simply will not join any platoon and continue on its own to
its destination.
The implementation of these mechanisms would not have
been possible without TraCI. On the one hand, the tool is
primarily used for controlling the vehicles in the simulation,
as well as identifying those vehicles that step onto induction
loops. On the other hand, it plays an important part in
extracting a set of metrics, regarding the number of requests
made and accepted in each simulation, in addition to fuel
consumption and CO2emissions. Since only a small subset
of SUMO ’s functionalities were put to use, TraCI’s usage is
proportionally limited in variety, even though its potential is
huge.
As mentioned before, different scenarios were devised, a
difference that resides essentially in the way a platoon and a
merging vehicle deal with the issue of lane merging. These
scenarios are described in detail in Section III
V. PRELIMINARY RESU LTS
Throughout the definition and development of each sce-
nario, experiments were carried out, in order to ensure its
proper implementation. The results that follow were obtained
by running each scenario 50 times and establishing a confi-
dence interval. A confidence level of 99% was used. Several
metrics were collected, namely, the average CO2emissions
per vehicle – in Kilograms –, the average fuel consumption per
vehicle – in Liters –, the average waiting time and average time
lost by not traveling at the desired velocity – both measured
in simulation ticks.
A. Reference Model
The reference model consists in a lane merging situation
without platooning or negotiation, where a vehicle attempts to
merge into a lane where traffic is flowing regularly.
TABLE III
MET RIC S OB SERV ED AS R ES ULT OF S IMU LATI ONS B ASE D ON T HE
REFERENCE MODEL.
Metric Interval
CO2Emissions (Kg) 0,321 ±0,009
Fuel Consumption (L) 0,138 ±0,004
Waiting time (ticks) 1,203 ±0,478
Time loss (ticks) 41,071 ±1,730
As expected, it was observed that vehicles traveling in the
acceleration lane stopped at the junction of the two lanes,
allowing for the passage of the vehicles circulating on the
main road, as per traffic rules.
Table III presents the results obtained for this reference
model for the various metrics considered.
B. Base Scenario
The base scenario consists in a lane merging situation
without negotiation, where a vehicle attempts to merge into
a lane where platoons transit.
TABLE IV
MET RIC S OB SERV ED AS R ES ULT OF S IMU LATI ONS B ASE D ON T HE BA SE
SC ENA RIO .
Metric Interval
CO2Emissions (Kg) 0.268 ±0.042
Fuel Consumption (L) 0.115 ±0.018
Waiting time (ticks) 6.952 ±1.220
Time loss (ticks) 94.599 ±1,730
Again, as expected, it was observed that vehicles traveling in
the acceleration lane stopped at the junction of the two lanes,
so as to allow for the passage of the platoons circulating on
the main road, as per traffic rules.
The results obtained for this scenario for the various metrics
considered are presented in Table IV .
C. No-Negotiation Scenario
In this scenario, the vehicles attempting to merge into a
platoon request permission from the platoon, to which the
platoon responds with a proposal, which, in turn, the vehicle
can only accept or decline.
TABLE V
MET RIC S OB SERV ED AS R ES ULT OF S IMU LATI ONS B ASE D ON T HE
SC ENA RIO O F NO NE GOT IATI ON.
Metric Interval
CO2Emissions (Kg) 0.185 ±0.005
Fuel Consumption (L) 0.080 ±0.002
Waiting time (ticks) 0±0
Time loss (ticks) 25.809 ±1.137
As expected, merging vehicles were accepted only when
the merging maneuver was physically possible. The merging
vehicles when accepted, joined the platoon, taking the most
convenient position within that platoon.
By analyzing the results in Table V and comparing these
results with those obtained for the reference model, it was
possible to observe a decrease of approximately 26% in both
the average amount of fuel consumed and the volume of CO2
emissions.
Additionally, the amount of time spent stopped or moving
very slowly decreased, as did the amount of time lost by
not traveling at the desired velocity, which decreased by
approximately 37%.
The improvement observed in the aforementioned metrics
stems from the fact that, by introducing an attempt at coor-
dinating the merging maneuver, by requesting the entry into
the incoming platoon, the need for slowing down or stopping
at the junction was limited to cases in which the vehicle is
denied entrance into the platoon.
D. One-To-One Negotiation Scenario
This scenario implies that the vehicle attempting to merge
into a platoon must request permission to enter the platoon,
by making a proposal, stipulating a value and desired position,
which develops into a negotiation between the vehicle and the
platoon. Once the platoon and the vehicle reach an agreement
the vehicle is allowed to enter the platoon or, if no agreement
is reached, the vehicle is denied entrance.
TABLE VI
MET RIC S OB SERV ED AS R ES ULT OF S IMU LATI ONS B ASE D ON T HE
SCENARIO OF ONE-TO-ONE NEGOTIATION.
Metric Interval
CO2Emissions (Kg) 0.178 ±0.008
Fuel Consumption (L) 0.077 ±0.005
Waiting time (s) 0.059 ±0.008
Time loss (ticks) 26.474 ±1.450
Comparing the results presented in Table VI with those
obtained in the previous scenario, a decrease of approximately
4% was observed in both the average amount of fuel consumed
and the volume of CO2emissions.
Across all metrics, there is a steep increase in performance
when comparing the reference model to both the take-or-leave
proposal and the naive-based negotiation scenarios.
VI. CONCLUSIONS
Lane merging can have a significant negative impact in road
network congestion levels and safety. Autonomous vehicles
enable faster and safer travel through the elimination of human
error as well as through leveraging the added processing
and mobile communication capabilities to negotiate and form
platoons of autonomous vehicles. One of the challenges posed
to the development of autonomous vehicles with platooning
capability concerns the management and negotiation of lane-
merging processes between platoons of autonomous vehicles
traveling on a road and autonomous vehicles entering onto
that road. With this work we provide an implementation
of the lane-merging maneuver in the SUMO simulator as it
may in the future be implemented in real-world scenarios.
Additionally we discussed two different mechanisms for pla-
toon formation: a take-or-leave proposal and a naive-based
negotiation. Furthermore, we lay out an agent-based design
for a framework in which additional negotiation techniques
may be implemented and tested with.
The simulation results obtained show platooning has an
important impact on reducing fuel consumption and CO2
emissions. Additionally the importance of using lane-merging
techniques and negotiation with platooning is highlighted
through the significant time loss incurred in the scenario where
no lane-merging techniques were applied to the simulation
when compared to a reference scenario with no platooning.
Platooning combined with lane-merging shows a significant
increase in performance in comparison to how traffic handles
itself in today’s road systems. Between the two lane merging
techniques that were analyzed no significant differences in
performance were detected. This lack of variation between
the metrics extracted and analyzed from the two techniques
may be attributed to two factors: a sub-optimal fine tuning
of the utility and cost functions of the bargaining method or
a possible similarity between the two negotiation techniques.
This assumption could be the theme of a more in-depth study
as future work.
Also as suggestions for further research, different and more
complex scenarios can and should be studied, including, but
not limited to, other variations to the negotiation process.
These other scenarios would include:
•Lane merging into a platoon with a negotiation process
in which the cost also determines the spot in the platoon
the vehicle will occupy;
•Lane merging into a platoon with slot assigning agents
in each intersection using machine learning to determine
the best cost to enter a platoon in that intersection;
•Platoon splitting into two different platoons by desired
speed, using clustering algorithms;
•Scenarios with more than one lane;
•Scenarios concerning different types of vehicles, with dis-
tinct characteristics, such as private cars, public transport,
and cargo transit;
•Scenarios using different road layouts.
Finally, our platform is part of a more complex framework
[14] that relies on the concept and techniques of Artifi-
cial Transportation Systems and Simulation resorting to the
metaphor of multi-agent systems [15]. Besides SUMO other
simulators are expected to be integrated in our platform,
allowing for the interoperability of multiple simulators through
the HLA standard and therefore more complex analyses can
be carried out. Following other examples of simulators inter-
operability [16], different simulation tools for vehicular ad-hoc
networks will be considered [17] so as to make the scenarios
more realistic in which the specificity and the influence of
vehicle communication approaches can be assessed as well.
REFERENCES
[1] P. Fernandes and U. Neves, “Platooning of Autonomous Vehicles with
Intervehicle Communication in SUMO Traffic Simulator. (English),” in
Annual Conference on Intelligent Transportation Systems, no. 13. IEEE,
September 2010.
[2] M. Michaelian and F. Browand, “Field Experiments Demonstrate Fuel
Savings for Close-Following.” in California Path Program Research
Report. University of Southern California, September 2000.
[3] L. S. Passos, R. J. F. Rossetti, and Z. Kokkinogenis, “Towards the
next-generation traffic simulation tools: a first appraisal,” in 6th Iberian
Conference on Information Systems and Technologies (CISTI 2011),
June 2011, pp. 1–6.
[4] B. C. da Silva, E. W. Basso, A. L. C. Bazzan, and P. M. Engel,
“Dealing with non-stationary environments using context detection,” in
Proceedings of the 23rd International Conference on Machine Learning,
ser. ICML ’06. New York, NY, USA: ACM, 2006, pp. 217–224.
[Online]. Available: http://doi.acm.org/10.1145/1143844.1143872
[5] M. Vasirani and S. Ossowski, “A market-based approach to accommo-
date user preferences in reservation-based traffic management,” in 6th
Workshop on Agents in Traffic and Transportation (ATT), Autonomous
Agents and Multiagent Systems (AAMAS), Toronto, Canada, May 2010.
[6] W. Yang, Z. Zhiyong, and Y. Jianhua, “Sequential game solution for
lane-merging conflict between autonomous vehicles,” in Inter. Conf. on
Intelligent Transportation Engineering. IEEE, 2016.
[7] M. Vasirani and S. Ossowski, “A market-inspired approach to
reservation-based urban road traffic management,” in 8th Int. Conf. on
Autonomous Agents and Multiagent Systems, ser. AAMAS’09, 2009, pp.
617–624.
[8] J. J. B. Vial, W. E. Devanny, D. Eppstein, and M. T. Goodrich,
“Scheduling autonomous vehicle platoons through an unregulated
intersection,” CoRR, vol. abs/1609.04512, 2016. [Online]. Available:
http://arxiv.org/abs/1609.04512
[9] W. Niu, Y.-q. Xu, J. Wu, and Y.-z. Tan, A Centralized Scheduling
Approach to Multi-Agent Coordination. Paris: Atlantis Press, 2015,
pp. 171–175.
[10] M. Amoozadeh, H. Deng, C.-N. Chuah, H. M. Zhang, and D. Ghosal,
“Platoon management with cooperative adaptive cruise control enabled
by VANET,” Vehicular Communications, vol. 2, no. 2, pp. 110 – 123,
2015.
[11] H. Rewald and O. Stursberg, “Cooperation of autonomous vehicles using
a hierarchy of auction-based and model-predictive control,” in 2016
IEEE Intelligent Vehicles Symposium (IV), 2016, pp. 1078–1084.
[12] G. Soares, J. Macedo, Z. Kokkinogenis, and R. J. F. Rossetti, “An
integrated framework for multi-agent traffic simulation using sumo and
jade,” in SUMO2013, The first SUMO user conference, 2013, pp. 15–17.
[13] G. Soares, Z. Kokkinogenis, J. L. Macedo, and R. J. F. Rossetti, “Agent-
based traffic simulation using sumo and jade: an integrated platform for
artificial transportation systems,” in Simulation of Urban MObility User
Conference. Springer, 2013, pp. 44–61.
[14] R. J. F. Rossetti, E. C. Oliveira, and A. L. Bazzan, “Towards a
specification of a framework for sustainable transportation analysis,” in
13th Portuguese Conference on Artificial Intelligence, EPIA, Guimar˜
aes,
Portugal. APPIA, 2007, pp. 179–190.
[15] R. J. F. Rossetti and R. Liu, “Advances in artificial transportation systems
and simulation,” 2014.
[16] J. Macedo, Z. Kokkinogenis, G. Soares, D. Perrotta, and R. J. F. Rossetti,
“A HLA-based multi-resolution approach to simulating electric vehicles
in simulink and sumo,” in 16th Int. IEEE Conference on Intelligent
Transportation Systems (ITSC 2013), Oct 2013, pp. 2367–2372.
[17] S. A. B. Mussa, M. Manaf, K. Z. Ghafoor, and Z. Doukha, “Simulation
tools for vehicular ad hoc networks: A comparison study and future
perspectives,” in 2015 International Conference on Wireless Networks
and Mobile Communications (WINCOM), Oct 2015, pp. 1–8.
