Cooperative Automated Driving: From Platooning to Maneuvering

Jeroen Ploeg

1,2 a

and Redmer de Haan

Cooperative Driving group, 2getthere B.V., Utrecht, The Netherlands

Department of Mechanical Engineering, Dynamics and Control Group,

Eindhoven University of Technology, Eindhoven, The Netherlands

Keywords:

Cooperative Automated Driving, Platooning, Maneuvering, Safety, Control System Architecture.

Abstract:

Cooperative automated driving (CAD) combines autonomous driving with cooperative driving, thereby yield-

ing a powerful approach to improve trafﬁc efﬁciency and safety. A very well-known example of CAD is

platooning. However, when extending this one-dimensional application to two-dimensional maneuvering,

covering a large number of trafﬁc scenarios while also including safety threats imposed by other trafﬁc or fail-

ing components of the automation system, a complex control system architecture may arise. To address this

challenge, an agent-based control system architecture is proposed employing explicit decision making. This

architecture is scalable with respect to the number of trafﬁc scenarios that can be handled, capable of including

safety features, and provides the ﬂexibility to adopt various controller design approaches at the same time.

1 INTRODUCTION

In recent years, autonomous driving has gained in-

creasing attention in the public press and in the sci-

entiﬁc community. Trafﬁc safety is the primary driver

for this development, but also other motivations ex-

ists, such as more effective use of the traveling time

and reducing the dependency on manpower.

Autonomous vehicles, however, do not intrin-

sically improve trafﬁc since they optimize towards

reaching their own goals. Cooperative driving, on the

other hand, aims for optimizing the collective behav-

ior, thus improving the trafﬁc system. Connectivity is

instrumental for cooperative driving because it allows

trafﬁc participants to share their intention easily and

precisely (de La Fortelle et al., 2014). When com-

bined with automation, a powerful approach arises to

improve trafﬁc safety and efﬁciency.

A well-known application of cooperative auto-

mated driving (CAD) is cooperative adaptive cruise

control (CACC) or platooning, which improves trafﬁc

throughput by adopting short intervehicle distances

(Ploeg et al., 2014). This is particularly of interest

in an automated transit network (ATN), i.e., a system

of automated people movers for ﬁrst-/last-mile public

transportation, in view of transport capacity. Truck

platooning is another promising application because

https://orcid.org/0000-0001-8332-5860

of the reduced aerodynamic drag at short distances

(Alam et al., 2015).

Next to ongoing developments in the ﬁeld of pla-

tooning, cooperative automated maneuvering attracts

attention to an increasing extent, acknowledging the

fact that trafﬁc is not a string of vehicles. Many ap-

proaches are still investigated in this ﬁeld. One such

approach relies on explicit decision making, which

was illustrated by i-GAME (Ploeg et al., 2018), a

European-funded project. Other projects, such as Au-

toNet2030, adopt an optimization-based approach for

path planning (Qian et al., 2016). A serious chal-

lenge for cooperative automated maneuvering, how-

ever, is posed by the fact that road trafﬁc involves

a large number of different scenarios, which are not

likely to be handled by a single integrated approach.

Moreover, next to nominal behavior, also safety mea-

sures come into play to handle failing system compo-

nents or emergency situations imposed by other traf-

ﬁc. This paper addresses this challenge by present-

ing a generic control architecture for CAD using an

agent-based approach, which intends to be scalable in

the sense that all possible trafﬁc scenarios can be in-

corporated without leading to a complicated control

system architecture while also being capable of in-

cluding safety-related features.

The next section ﬁrst presents a brief summary of

developments in controller design for platooning and

the emerging ﬁeld of cooperative automated maneu-

Ploeg, J. and de Haan, R.

Cooperative Automated Driving: From Platooning to Maneuvering.

DOI: 10.5220/0008346300050010

In Proceedings of the 5th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2019), pages 5-10

ISBN: 978-989-758-374-2

i ‒ 1

i + 1

i ‒ 1

i + 1

i ‒ 2

Figure 1: Platoon of ATN vehicles.

vering. Next, Section 3 focuses on safety of CAD.

Section 4 proposes a generic control architecture after

which the main results are summarized in Section 5.

2 COOPERATIVE AUTOMATED

DRIVING

A very well-known CAD application, focusing on

longitudinal automation, is vehicle platooning, the

main aspects of which will be brieﬂy summarized in

Section 2.1. When also taking lateral vehicle motion

into account, the concept of platooning needs to be

extended towards cooperative automated maneuver-

ing, an example of which is presented in Section 2.2.

2.1 Platooning

An example platooning set-up is depicted in Fig. 1,

where v

is the speed of the vehicle with index i

and d

is the intervehicle distance between vehicle

i and the downstream (forward) vehicle i − 1. The

main control objective is to regulate d

to a desired

value d

r,i

, to which end, in this example, a one-

vehicle look-ahead communication topology is em-

ployed next to on-board sensors, such as forward-

looking radar and/or camera, to measure the interve-

hicle distance and the range rate. Note that the pla-

tooning controller is known as cooperative adaptive

cruise control (CACC), since it can be viewed as an

extension of adaptive cruise control (ACC) with wire-

less vehicle-to-vehicle (V2V) communications.

An important requirement for platooning is known

as string stability (Ploeg et al., 2014), i.e., the attenu-

ation of the effects of disturbances along the string

in upstream direction. This requirement is usually

formalized by requiring that the L

signal norm (en-

ergy) or the L

∞

signal norm (amplitude) of the ve-

locity v

or acceleration a

does not amplify for in-

creasing i. Adopting a constant time-gap spacing pol-

icy, i.e., d

r,i

(t) = r + hv

(t) at time t, where r is the

standstill distance and h the time gap, is beneﬁcial for

string-stable platoon behavior. In this case, a mini-

mum time gap h

min

exists above which string stability

can be guaranteed. But to also obtain string stability at

short intervehicle distances (h ≈ 0.3 s), wireless V2V

communication is required. In the example of Fig. 1,

taken from (Ploeg et al., 2014), the input u

(desired

acceleration) of vehicle i is communicated to the up-

stream vehicle i + 1, which can lead to string stable

time gaps as low as h

min

= 0.24 s. Since u

cannot be

measured by the on-board sensors of the downstream

vehicle, it must be communicated, hence the need for

wireless V2V communication.

Many platooning controllers have been proposed

in literature, see (Ploeg et al., 2014) and the references

contained therein, some of them employing more

complex communication topologies or even varying

topologies (Santini et al., 2019). Despite this vast

amount of literature, however, some challenges still

remain, among which the control of heterogeneous

vehicle platoons and, even more important, the de-

sign of safety measures in the case of, e.g., sudden

packet loss of the V2V link. Nevertheless, trafﬁc is

certainly not limited to platooning scenarios, which is

why the ﬁeld of CAD is extended towards cooperative

automated maneuvering, as illustrated next.

2.2 Maneuvering

Automated crossing of an intersection without traf-

ﬁc lights is a good example of cooperative automated

maneuvering. This particular application received

quite some attention in literature, see, e.g., (Morales

Medina et al., 2018). In this section, however, we fo-

cus on another example that is very illustrative for the

upcoming architecture proposal, being a highway lane

reduction, involving zipping of two vehicle platoons,

as presented earlier in (Ploeg et al., 2018).

Consider a platoon L on the left lane, with mem-

bers L

, i = 1, . . . , m, and a platoon R with members

, j = 1, . . . , n, on the right lane, as illustrated in

Fig. 2 for m = 3 and n = 4. The lane-reduction sce-

nario can then be solved by the following sequence of

maneuvers, initiated by an interaction protocol that is

implemented through wireless V2V communications.

1. Pair-up R2L — The ﬁrst phase entails sending

merge requests by the vehicles in L to the ones

in R. Next, each vehicle R

ﬁnds an appropriate

merging partner L

to merge in front of R

, thus

creating pairs {L

, R

} using a V2V handshaking

mechanism. The actual maneuver is that the vehi-

cles in R slow down to create an appropriate dis-

tance towards their merging partner in L, which is

implemented by R

activating a CACC controller

with L

as target vehicle. Since it may also happen

that the preceding in-lane vehicle R

j−1

brakes for

some reason, vehicle R

also activates a ‘separa-

tion controller’, guaranteeing a certain minimum

distance towards R

j−1

. This procedure is executed

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

L R

Pair-up R2L Pair-up L2R Gap ready Merged

L R L R L R

Figure 2: Phases of the lane-reduction scenario.

for all vehicles simultaneously.

2. Pair-up L2R — In the next phase, the same type

of procedure is followed but in opposite direction:

Each vehicle L

ﬁnds a appropriate merging part-

ner in R, which usually will be R

j−1

given the pair

, R

} from the previous phase, and activates its

CACC controller with R

j−1

as target, while also

executing a separation controller with L

i−1

as tar-

get vehicle. This procedure, however, is executed

sequentially in upstream direction to prevent large

decelerations of the vehicles in the tail of the pla-

toon due to the gap-making maneuver.

3. Gap ready — When the gap is large enough, vehi-

cle R

signals its pair L

that it is allowed to initiate

the actual merge maneuver. Only when vehicle L

starts to change lanes, phase 2 is initiated for the

next vehicle L

i+1

From the above description, it is clear that this ap-

proach is based on explicit decision making, driven by

the interaction protocol, which initiates a maneuver

sequence. Each maneuver is executed by one or more

controllers which have a simple objective, such as

regulating a desired distance or a minimum distance,

or making a lane change. Note that other approaches

exist that do not rely on explicit decision making, such

as the Model Predictive Control approach presented

in (Qian et al., 2016). This would, however, require

all vehicles to have the exact same type of controller,

which might not be feasible given the fact that there

are various vehicle manufacturers.

Operational

safety

Functional

safety

(ISO 26262)

SOTIF

(ISO/PAS

21448)

Behavioral

safety

(emerging)

Best practices

(guidelines,

experience)

Figure 3: Aspects of operational safety.

Until now, only nominal behavior has been con-

sidered. In case of practical deployment, however,

also safety comes into play, as explained next.

3 ROAD SAFETY

Practical deployment of CAD applications requires

a structured approach to road safety. This section

brieﬂy summarizes some important types of safety

and presents relevant threats in the scope of platoon-

ing, thereby motivating that additional vehicle con-

trollers are required to ensure safe behavior.

3.1 Standardization

Operational safety, which is used here as an umbrella

term for all types of safety, involves both ‘safety of

the intended functionality’ (SOTIF) and functional

safety, as depicted in Fig. 3. Here, SOTIF refers to

the ability of the system to correctly comprehend the

environmental situation and respond safely by activat-

ing appropriate countermeasures. SOTIF is recently

standardized as ISO 21448 (ISO/PAS 21448, 2019).

Functional safety, on the other hand, is the absence

of unreasonable risk due to hazards caused by mal-

functioning behavior of subsystems of the automated

vehicle, as standardized in the notorious ISO 26262

(ISO 26262-1, 2018). It should be mentioned that

ISO 26262 actually does not cover fully automated

road vehicles. Instead, this standard is limited to par-

tial automation, as implemented by advanced driver

assistance systems (ADAS), among which ACC.

Next to SOTIF and functional safety, also the no-

tion of behavioral safety has been recently introduced

as “an aspect of system safety that focuses on how

a system should behave normally in its environment

to avoid hazards and reduce the risk of mishaps”

(Waymo, 2017). Behavioral safety thus refers to

whether the programmed response of an automated

vehicle to common trafﬁc situations is safe.

Finally, best practices are still important for the

development of automated vehicles, mainly due to the

limited scope of ISO 26262.

Cooperative Automated Driving: From Platooning to Maneuvering

Table 1: Platoon-speciﬁc threats in the scope of ‘safety of

the intended functionality’ (SOTIF).

ID Threat Countermeasure

SO1 Emergency

brake of

equipped

vehicle

No speciﬁc countermeasure is

required in this case, provided the

platoon is string stable.

SO2 Intermittent

V2V packet

loss

Graceful degradation — Upon

exceeding a packet loss threshold,

a smooth switch from CACC to

ACC is performed, while increas-

ing the following distance to a

safe and string stable value.

SO3 Unequipped

in-lane vehicle

Graceful degradation — The

same countermeasure as in SO2

applies.

SO4 Emergency

brake of

unequipped

in-lane vehicle

Fail safety — A collision avoid-

ance mechanism must be acti-

vated while messaging all up-

stream platoon members, allow-

ing those to respond in a timely

manner.

SO5 Cut-in/-

through of

unequipped

vehicle

Graceful degradation — In most

cases, this threat requires a similar

response as in SO3; However,

if the alien vehicle signiﬁcantly

decelerates at the same time, a

fail safety mechanism must be

activated, as in SO4.

3.2 CAD Safety Threats

SOTIF and functional safety are the main types of

safety to take into account when developing CAD

systems since these are standardized. This section

lists common threats, related to SOTIF and functional

safety, for CAD systems in general and platoons in

particular. To this end, Table 1 ﬁrst summarizes some

important threats in the scope of SOTIF. In this table,

an ‘equipped vehicle’ refers to an automated vehicle

with wireless communication capability.

As can be clearly seen from this table, all listed

threats relate to dangerous situations imposed by

other trafﬁc (SO1 and S03–SO5) or to inherent limi-

tations of the automated vehicle’s environmental per-

ception sensor suite, in particular the V2V commu-

nication (SO2). In other words, SOTIF encompasses

threats that inherently exist under normal conditions

while driving in mixed trafﬁc.

Functional safety exclusively focuses on compo-

nent failures. Some important failures, particularly

related to platooning, are listed in Table 2. This con-

cerns persistent packet loss of the V2X communi-

cation system (as compared to ‘normal’ packet loss,

which is covered by SOTIF). Malfunctioning behav-

ior of on-board environmental sensors is also consid-

ered, assuming that failure of the environmental per-

Table 2: Platoon-speciﬁc failures in the scope of functional

safety.

ID Failure Countermeasure

FS1 Persistent V2V

packet loss

Graceful degradation — With

the on-board EPS still fully func-

tional, V2V failure is counter-

acted by smoothly switching from

CACC to ACC, including increase

of the following distance to regain

safety and string stability.

FS2 EPS failure Fail safety — It is technically pos-

sible but unsafe to continue pla-

tooning using V2V only, because

unequipped vehicles or other ob-

jects can no longer be detected.

Therefore, EPS failure triggers a

collision avoidance mechanism as

the default fail-safety measure.

FS3 Failure of

preceding

equipped

vehicle

Fail safety — This is a combina-

tion of an emergency stop of the

preceding vehicle and FS1, which

can only be treated as a fail-safety

situation (collision avoidance)

while messaging all upstream pla-

toon vehicles, allowing those to

respond in a timely manner.

ception system (EPS) can be detected, either directly

or through comparison with redundant on-board sen-

sors. Finally, an equipped vehicle may be subject to

a major failure, due to which the vehicle performs an

emergency stop and, at the same time, all systems are

shut down, among which the wireless communication

system. The latter type of threat is particularly rele-

vant for platoons of people movers, which typically

perform an emergency stop when essential systems

exhibit malfunctioning behavior.

Both table Table 1 and Table 2 also show possi-

ble countermeasures for each threat or failure, cate-

gorized as either graceful degradation or fail safety.

Consequently, in addition to the nominal controllers

mentioned in Section 2, controllers are needed for fail

safety, e.g., a collision-avoidance controller, and for

graceful degradation. An example of the latter is au-

tomatically reverting from CACC to ACC while in-

creasing the intervehicle distance in the case of per-

sistent V2V packet loss.

4 CONTROL SYSTEM

ARCHITECTURE

Section 2 concerned controller design for nominal be-

havior, whereas Section 3 touched upon non-nominal

situations. To automatically control vehicles that col-

laboratively execute various trafﬁc scenarios under

both nominal and non-nominal conditions, a layered

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems

Coordination layer

Navigation layer

Motion control layer

Figure 4: Layered architecture of CAD systems (dashed ar-

rows indicate information exchange through wireless com-

munications; the white vehicle is unequipped).

control system architecture is proposed, inspired by

(Horowitz and Varaiya, 2000), among others.

4.1 A Layered Software Architecture

Three main control levels can be distinguished in the

scope of CAD, as summarized below and visualized

in Fig. 4.

• The centralized navigation layer is responsible for

scheduling and routing, taking into account fuel

consumption and travel time, among others. In

case of truck platooning, this layer primarily in-

volves logistics, whereas in the case of ATNs, it

would focus on ﬂeet control while also keeping

track of vehicle status and maintenance schedules.

• The intermediate coordination layer is responsi-

ble for coordination among the vehicles in a coop-

erative maneuver. This layer may exclusively in-

volve decision making, hence executing the afore-

mentioned interaction protocol, but may also act

as a higher-level feedback control loop. An ex-

ample of the latter is presented in (Zegers et al.,

2017), concerning the design of a mechanism to

guarantee platoon coherency subject to velocity

constraints. This layer’s implementation should

be distributed to support the distributed nature of

many trafﬁc maneuvers.

• At the individual vehicle level, the motion con-

trol layer performs the actual real-time control of

the automated vehicle in order to execute the re-

quired maneuvers. Consequently, this layer in-

volves controllers for longitudinal vehicle motion,

e.g., (C)ACC, and lateral motion, such as lane

keeping.

This control system hierarchy is very similar to

the three levels commonly distinguished for the hu-

man driving task (Michon, 1985), being the strategic

level, the tactical level, and the operational level, re-

spectively, which were the terms used in (Ploeg et al.,

2018). The main motivation for this layered architec-

ture is twofold: First, it supports the explicit inclusion

scenario 1:

maneuver sequencing

maneuver 1:

agent activation

maneuver m:

agent activation

maneuver M :

agent activation

scenario N :

maneuver sequencing

ready

agent 1:

control

agent i:

control

agent j:

control

agent K:

control

activate

neg.

negotiate

Figure 5: The coordination layer (light gray) and the mo-

tion control layer (dark grey) in an agent-based control ap-

proach.

of interaction protocols, and second, in the motion

control layer, there is freedom to adopt various con-

troller design approaches. The latter is particularly

relevant in view of the different road vehicle brands.

4.2 Agent-based Control

To further detail the proposed architecture, in par-

ticular the coordination layer and the motion control

layer, one could think of road trafﬁc as a set of sce-

narios. Each scenario is built from (a sequence of)

maneuvers, which are executed by one or more con-

trollers, or agents, having a simple objective such as

speed control, distance control, etc.. If more than one

agent is required to execute the maneuver, the agents

can ‘negotiate’ among each other about which one ac-

tually controls the vehicle motion.

Taking this simple road trafﬁc ontology as a basis,

the coordination layer is then responsible for execu-

tion of a scenario by subsequently activating the re-

quired maneuvers. Each maneuver, in turn, is imple-

mented by one or more agents for the longitudinal and

lateral vehicle motion. This agent-based control ap-

proach, which has the advantage of being ﬂexible and

computationally non-demanding (Jennings and Buss-

mann, 2003), is depicted in Fig. 5. Note that this

approach is very similar to that of hybrid automata

(Huang et al., 2019).

Consider the lane-reduction scenario as discussed

in Section 2.2 to illustrate this concept. This scenario

requires the right-lane vehicles to make a gap for the

left-lane vehicles, i.e., a gap-making maneuver. Next,

the left-lane vehicles need to perform a lane-change

maneuver, and the scenario ends with all vehicles on

the right lane performing a vehicle-following ‘maneu-

ver’. The sequence of these maneuvers is controlled

by the interaction protocol, which runs in the coordi-

nation layer. The gap-making maneuver entails two

control objectives: realizing a desired distance to-

Cooperative Automated Driving: From Platooning to Maneuvering

wards the merging vehicle, while guaranteeing a min-

imum distance towards the preceding in-lane vehicle.

Hence, two agents are involved in executing this ma-

neuver: a CACC agent to regulate the distance to-

wards the merging vehicle, and a separation agent to

guarantee a minimum distance towards the preceding

in-lane vehicle. The lane change is performed by a

lane-change agent, while the ﬁnal vehicle-following

situation is realized through the CACC agents of all

vehicles. Negotiation among agents takes place dur-

ing the gap-making maneuver, since the separation

agent must have priority above the CACC agent in

case the preceding in-lane vehicle brakes; likewise,

the CACC agent has priority if the preceding in-lane

vehicle decides to accelerate for some reason.

During all maneuvers, it may be required to also

activate a collision avoidance agent as a fail safety

measure, capable of performing an emergency stop in

case dangerous situations occur during the scenario

execution, thus overruling other active agents. In ad-

dition, an ACC agent might take over from the CACC

agents in case of packet loss, thus implementing a

graceful degradation measure.

5 CONCLUSION

It was argued that cooperative automated driving re-

gards road trafﬁc as a system instead of individual

vehicles, thus having the potential to improve traf-

ﬁc efﬁciency and safety. Platooning is a well-known

example in this ﬁeld, but must be extended in two

directions: First, to cover multiple trafﬁc scenarios,

one-dimensional platooning must evolve into two-

dimensional maneuvering and second, practical de-

ployment requires inclusion of safety measures. To

this end, a software architecture for the control sys-

tem was proposed utilizing an agent-based approach.

This architecture will be implemented in the near fu-

ture to realize cooperative behavior in a ﬂeet of people

movers.

REFERENCES

Alam, A., Besselink, B., Turri, V., M

artensson, J., and Jo-

hansson, K. H. (2015). Heavy-duty vehicle platooning

for sustainable freight transportation: A cooperative

method to enhance safety and efﬁciency. IEEE Con-

trol Systems Magazine, 35(6):34–56.

de La Fortelle, A., Qian, X., Diemer, S., Gr

egoire, J.,

Moutarde, F., Bonnabel, S., Marjovi, A., Martinoli,

A., Llatser, I., Festag, A., and Sj

oberg, K. (2014).

Network of automated vehicles: The AutoNet 2030

vision. In Proceedings of the 21st World Congress on

Intelligent Transport Systems.

Horowitz, R. and Varaiya, P. (2000). Control design of an

automated highway system. Proceedings of the IEEE,

88(7):913–925.

Huang, Z., Chu, D., Wu, C., and He, Y. (2019). Path plan-

ning and cooperative control for automated vehicle

platoon using hybrid automata. IEEE Transactions on

Intelligent Transportation Systems, 20(3):959–974.

ISO 26262-1 (2018). Road vehicles – Functional safety –

Part 1: Vocabulary. Technical report, International Or-

ganization for Standardization, Geneva, Switzerland.

ISO/PAS 21448 (2019). Road vehicles – Safety of the in-

tended functionality. Technical report, International

Organization for Standardization, Geneva, Switzer-

land.

Jennings, N. R. and Bussmann, S. (2003). Agent-based

control systems: Why are they suited to engineering

complex systems? IEEE Control Systems Magazine,

23(3):61–73.

Michon, J. A. (1985). A critical view of driver behavior

models: What do we know, what should we do? In

Evans, L. and Schwing, R. C., editors, Human Be-

havior and Trafﬁc Safety, pages 485–520. Springer,

Boston, MA.

Morales Medina, A. I., van de Wouw, N., and Nijmeijer,

H. (2018). Cooperative intersection control based on

virtual platooning. IEEE Transactions on Intelligent

Transportation Systems, 19(6):1727–1740.

Ploeg, J., Semsar-Kazerooni, E., Morales Medina, A. I., de

Jongh, J. F. C. M., van de Sluis, J., Voronov, A., En-

glund, C., Bril, R. J., Salunkhe, H., Arr

ue, A., Ru-

ano, A., Garc

ıa-Sol, L., van Nunen, E., and van de

Wouw, N. (2018). Cooperative automated maneuver-

ing at the 2016 Grand Cooperative Driving Challenge.

IEEE Transactions on Intelligent Transportation Sys-

tems, 19(4):1213–1226.

Ploeg, J., van de Wouw, N., and Nijmeijer, H. (2014). L

string stability of cascaded systems: Application to

vehicle platooning. IEEE Transactions on Control

Systems Technology, 22(2):786–793.

Qian, X., de La Fortelle, A., and Moutarde, F. (2016). A

hierarchical Model Predictive Control framework for

on-road formation control of autonomous vehicles.

In Proceedings of the 2016 IEEE Intelligent Vehicles

Symposium, pages 376–381.

Santini, S., Salvi, A., Valente, A. S., Pescap

e, A., Segata,

M., and Cigno, R. L. (2019). Platooning maneuvers

in vehicular networks: A distributed and consensus-

based approach. IEEE Transactions on Intelligent Ve-

hicles, 4(1):59–72.

Waymo (2017). On the road to fully self-driving: Waymo

safety report.

Zegers, J. C., Semsar-Kazerooni, E., Fusco, M., and Ploeg,

J. (2017). A multi-layer control approach to truck pla-

tooning: Platoon cohesion subject to dynamical limi-

tations. In Proceedings of the 5th IEEE International

Conference on Models and Technologies for Intelli-

gent Transportation Systems, pages 128–133.

VEHITS 2019 - 5th International Conference on Vehicle Technology and Intelligent Transport Systems