Collaborative and Distributed QoS Monitoring and Prediction:

A Heterogeneous Link Layer Concept towards always Resilient V2X

Communication

Daniel Hincapie, Ahmad Saad and Josef Jiru

Fraunhofer Institute for Embedded Systems and Communication Technologies, Munich, Germany

Keywords:

Access Network Selection, Quality of Service, Prediction, Monitoring, Resilient Communication, Au-

tonomous Driving, Vehicle-To-X Communication.

Abstract:

Vehicle-To-Everything (V2X) communication is a fundamental pillar of autonomous driving. It enables the

exchange of safety-critical data between vehicles, infrastructure and pedestrians to enhance the awareness of

the surrounding environment and coordinate the execution of collective functionalities vital to achieve full

automation. Due to the safety-critical nature of the interchanged information, V2X communication must be

resilient, so that it provides reliable connectivity despite of the very dynamic characteristics of both its environ-

ment and network topology. In this position paper, we propose a novel concept that aims at achieving resilient

V2X communication. We introduce the Quality of Service Manager (QoSM), a collaborative and distributed

implementation concept for the Heterogeneous Link Layer (HLL) that operates on the top of the Medium

Access Control (MAC). The QoSM ﬁrst monitors and predicts QoS indicators of Radio Access Technologies

(RAT) in Heterogeneous Vehicular Networks (HetVNET). Then, it determines, under the principle of Always

Resiliently Connected (ARC), and sets timely the conﬁguration settings of RAT that meet performance and

reliability requirements of autonomous driving applications. Should it not be possible to fulﬁll applications

demands, the QoSM can instruct applications in advance to lower the requirements or run in a safe mode.

Like in many autonomous driving applications, the concept of our proposed QoSM is distributed and collab-

orative to enhance accuracy, self-awareness and safety. QoSMs shall be grouped hierarchically according to

correlation of applications demands, conditions of communication links and mobility information. Group’s

members share monitored and predicted indicators, as well as conﬁguration settings. This information is used

to determine collectively the conﬁguration of the HetVNET. On the one hand, sharing information among

QoSMs increases the amount of correlated data used by prediction algorithms, which improves prediction

accuracy. On the other hand, hierarchical groups allow to extend the proposed methodology to other hierar-

chical elements of the access and core network. With this position paper, we intend to open the discussion on

the importance of implementing protocols for sharing parameters that allow distributed and collaborative QoS

management for resilient V2X communication.

1 INTRODUCTION

Autonomous driving systems rely on collective and

safety-critical applications that must meet stringent

performance and reliability requirements. Thanks

to the rapid development of wireless communication

technologies, Vehicle-To-Everything (V2X) commu-

nication is expected to boost the development of au-

tonomous driving by enhancing the detection of sur-

rounding conditions and allowing coordinating the

execution of collective functionalities (Zheng et al.,

2015b; Zheng et al., 2015a). However, provisioning

satisfactory performance in V2X communication is a

difﬁcult task for wireless access networks due to the

mobility of involved entities, i.e., vehicles, infrastruc-

ture and pedestrians, and dynamic change of network

topology (Zheng et al., 2015b). To counteract this

issue, entities are equipped with Heterogeneous Ve-

hicular NETworks (HetVNET) that integrate different

Radio Access Technologies (RAT) such as Long Term

Evolution (LTE) and Dedicated Short Range Commu-

nication (DSRC). Communicating entities must then

select the interface that best suits running applications

needs in a decision process known as Access Network

Selection (ANS).

Hincapie, D., Saad, A. and Jiru, J.

Collaborative and Distributed QoS Monitoring and Prediction: A Heterogeneous Link Layer Concept towards always Resilient V2X Communication.

DOI: 10.5220/0007801706010608

In Proceedings of the 5th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2019), pages 601-608

ISBN: 978-989-758-374-2

601

1.1 HetVNET

Having technology diversity, i.e., different RAT, aims

to provide communicating entities with complemen-

tary technologies. Mobile cellular networks such as

LTE can provide wide geographical coverage, but

cannot efﬁciently support real-time information ex-

change for local areas. Conversely, DSRC is de-

signed for short-range communications and supports

well real-time safety messages distribution in a very

limited service area (Zheng et al., 2015a). Technol-

ogy diversity also offers alternative connectivity paths

when the degradation of a technology threatens appli-

cations functionalities.

Selecting effectively the access technology that

fulﬁlls the performance of running applications ac-

cessing the HetVNET is not trivial. From the tech-

nological point of view, RAT performance depends

on numerous radio and network resources, so the

universe of variables that have to be taken into ac-

count by ANS algorithms is large and tends to in-

crease in amount and complexity for the coming 5G

technologies. Regarding the applications accessing

the HetVNET, they must remain agnostic of the het-

erogeneity of the underlying infrastructure driven by

vertical handover procedures that offer service conti-

nuity, robustness/availability, and consistency main-

tained transparently (Gustafsson and Jonsson, 2003;

Louta and Bellavista, 2013).

Whereas handover procedures have received spe-

cial attention in fourth Generation (4G)-related re-

search and several standardization efforts provide a

framework for seamless vertical handover, decision-

making methods have not been standardized (Louta

and Bellavista, 2013). Nevertheless, ANS solutions

can be broadly divided into two categories: user-

and network-initiated algorithms. In user-initiated

schemes, no cooperation between technologies of

the Heterogeneous Networks (HetNet) is considered,

so entities individually take selﬁsh RAT selection

decisions to maximize individual utility functions,

which may not provide a globally optimum solution.

Conversely, network-initiated (also called network-

centralized) ANS algorithms seek to optimize global

network performance parameters, since decisions are

taken by a centralized controller that has an overall

view of the network (Roy et al., 2018). However,

there are downsides to centralized network-controlled

HetNet derived from the fact that information gath-

ering and decision tasks are not distributed. Chief

among these and very critical for V2X communica-

tion are the issues of timeliness of switching, scal-

ability of the system to multiple clients with multi-

ple interfaces, and the issue of how to obtain global

control of client-RAT association (Wang et al., 2016).

Representative vertical handover schemes with em-

phasis on both user- and network-initiated decision

process for ANS are presented and analyzed in (Louta

et al., 2011; Wang et al., 2016; Mohamed et al., 2012;

Lahby et al., 2015; Trestian et al., 2012; Wang et al.,

2016).

Independently of the scheme chosen to select the

access technology, decision-making methods must

deﬁne criteria to quantify and evaluate RAT perfor-

mance. We now take a look at the Always Best Con-

nected (ABC) concept, a decision criterion commonly

employed in ANS algorithms for HetVNET.

1.2 Selection Criteria

The ABC concept suggests that entities shall be not

only always connected, but also connected through

the best available technology at all times (Gustafs-

son and Jonsson, 2003). In line with the ABC

principle, but also lack of unity regarding the char-

acteristics and parameters used as decision crite-

ria, selection algorithms rely on performance indi-

cators of RAT that they monitor periodically (Louta

and Bellavista, 2013). Performance indicators can

be, but not limited to: Link quality measurements

such as Signal-to-Noise Ratio (SNR), Received Sig-

nal Strength (RSS), and Carrier-to-Interference-Ratio

(CIR); network-related indicators considering cover-

age, bandwidth availability and load; and Quality of

Service (QoS) aspects such as throughput, latency, jit-

ter and Packet Delivery Rate (PDR). They all or a

sub-set of them are retrieved from available RAT and

analyzed to select the access that outperforms and of-

fers the best connectivity (Louta and Bellavista, 2013;

Zheng et al., 2015a). After deciding which interface

should be used to transmit data and when it should be

activated, handover involving seamless transition to

the new network point of attachment has to take place

(Louta and Bellavista, 2013).

Resilience describes an autonomous system that

continues to function reliably despite of the occur-

rence of expected or unexpected changes (Fraunhofer

ESK Institute, 2018). Wireless technologies play

a crucial role in autonomous driving (Zheng et al.,

2015b), particularly, because its safety-critical appli-

cations require ultra-reliable V2X connectivity able to

at least ensure minimum performance requirements

for their functionalities. In case those requirements

cannot be met under extreme degradationof RAT con-

ditions, a mechanism should alert safety-critical ap-

plications in advance, so that they start failure pro-

cedures and modes that guarantee basic functionali-

ties. Hence, ANS in HetNets for V2X communica-

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

602

tion should not only aim to obtain high performance

in terms of link, network and QoS indicators, but also

to achieve resilient connectivity. To the best of our

knowledge, there does not exist any ANS algorithm

that takes account of resilience as decision criterion.

1.3 Contribution of this Position Paper

Towards the goal of supporting the dynamic and in-

stant composition of HetVNET, the work in (Zheng

et al., 2015a) introduced the Heterogeneous Link

Layer (HLL). The HLL operates on the top of

Medium Access Control (MAC) layer and enables

global managementof network resources to meet QoS

requirements of safety/non-safety services. However,

a uniﬁed approach to enable cooperation among mul-

tiple systems is difﬁcult to achieve due to the huge

amount of radio resources and unique characteris-

tics of RAT. In this position paper, we propose a

novel HLL concept that aims to achieve resilient ap-

plications: the Quality of Service Manager (QoSM).

Since our purposed QoSM is oriented to achieve re-

silience, we introduce the concept Always Resiliently

Connected (ARC) (an adaptation of the ABC vision

(Gustafsson and Jonsson, 2003)) according to which

applications accessing RAT of HetVNET must always

be connected to the access technology that provides

the most resilient conﬁguration proﬁle. A conﬁgura-

tion proﬁle is the object containing all general techno-

logical resources offered by RAT and target QoS in-

dicators. Resources may include features such as car-

rier aggregation, dual-connectivity, or methodologies

such as trafﬁc steering, packet duplication, among

others; target QoS indicators on the other hand, in-

dicate the values that are intended to be achieved

by conﬁguring the corresponding features. This ap-

proach allows the QoSM to address the problem of

uniﬁed management by grouping the particularities

of RAT in a more abstract concept: the conﬁguration

proﬁle. Consequently, according to our approach, the

problem of selecting an access technology under the

ABC principle turns into the problem of selecting the

most resilient conﬁguration proﬁle. In regards of eval-

uation, we deﬁne, conceptually based on link qual-

ity, network status and QoS indicators, three quanti-

ﬁers of conﬁguration proﬁles resilience: Predictabil-

ity, stability, and ﬂexibility. To achieve ARC applica-

tions, the proposed approach is both collaborativeand

distributed. This enables functional grouping to coun-

teract the scalability and timeless problems of central-

ized solutions, as well as the lack of global overview

of entity-based approaches.

This paper is organized as follows. Section 2 de-

ﬁnes the indicators used to qualify and quantify re-

silience. Section 3 describes the concept, character-

istics and tasks of the QoSM. Section 4 presents our

distributed and collaborative monitoring and predic-

tion approach, and how it can be scaled to improve

the QoS performance of higher hierarchical elements

of the access network. Concluding remarks and future

work are given in Section 5.

2 RESILIENCE INDICATORS

In order to determine the characteristics that shall

rule Conﬁguration Proﬁle Selection (CPS) for ARC

safety-critical application, it is necessary to deﬁne the

resilience indicators to be evaluated by the decision

algorithm. Resilience of systems is determined by

numerous components: memory, robustness, redun-

dancy, resourcefulness, and the capacity to recover

quickly from failures and adapt to them (Connelly

et al., 2017; Linkov, 2018). However, resilience fea-

tures can not be straightforwardly used as decision

criteria for CPS, because they are abstract concepts

on which there is not general consensus to quantify

them.

Based on the aforementioned characteristics and

with the purpose of quantifying them, we deﬁne (for

the moment conceptually) three resilience indicators:

Predictability, Stability and Flexibility.

2.1 Predictability

Predictability quantiﬁes, for prediction horizon, the

prediction accuracy of the QoS indicators that result

from selecting a given conﬁguration proﬁle. This fea-

ture allows to rank conﬁguration proﬁles according to

the certainty about their future performance. A pre-

dictability quantiﬁer could be, for example, the Mean

Squared Prediction Error (MSPE) of throughput or

latency. For some latency-sensitive applications, it

would be important to choose the conﬁguration pro-

ﬁle that lasts the longest in order to avoid latency

caused by hardware reconﬁguration and overheads

accessing the network. In that case, an approach to

quantify this resilience indicator would be to ﬁx a

maximum MSPE for a the set of QoS indicators (those

of interest for the application), and then evaluate the

maximum prediction horizon that can be achieved by

the conﬁguration proﬁles without exceeding the max-

imum MSPE value.

2.2 Stability

This concept intends to represent the capacity of

a conﬁguration proﬁle to keep QoS indicators in a

Collaborative and Distributed QoS Monitoring and Prediction: A Heterogeneous Link Layer Concept towards always Resilient V2X

Communication

603

steady state or regime. A conﬁguration proﬁle shall

be able to absorb changes in conditions to a certain

extent (Connelly et al., 2017). If a disruptive condi-

tion, e.g., an interferer, perpetuates changes in con-

ditions that exceed some intrinsic tolerance threshold

of the conﬁguration proﬁle, e.g., minimum SNR, the

proﬁle must adopt a regime where the resources of the

proﬁle are fundamentally different. Some quantiﬁers

of this indicator could be: jitter, gradient of through-

put or latency with respect to SNR, and covariance

between QoS and link quality indicators.

2.3 Flexibility

Stability may not be sufﬁcient to keep functionality of

a system. A conﬁguration proﬁle may be very stable

and sustain its regime against one type of disturber,

but it may fail when the nature of the disturbance

changes (Connelly et al., 2017). For example, using

a conﬁguration proﬁle that implements interleaving

is effective to preserve the PDR in environments ex-

posed to impulsive noise. The same proﬁle can vary

the interleaving depth according to the impulse width

to sustain the targeted PDR, i.e., it is a stable conﬁg-

uration proﬁle for PDR (but maybe not for latency).

However, if the noise source is not impulsive but con-

stant, such conﬁguration proﬁle may experiment low

PDR, i.e., it is considered as poorly ﬂexible. In con-

trast with stability, in which the nature of agents is

invariant and only their characteristics (e.g., magni-

tude or duration) vary, ﬂexibility measures the pro-

ﬁle capacity to maintain the value of QoS indicators

when the environment conditions change fundamen-

tally. Indicators of ﬂexibility can be derived from the

same stability quantiﬁers when they are obtained un-

der well known different scenarios, e.g., noise sources

interfering different frequency bands.

Service

QoS

Manager

Application

RF interfaces

RF transmission

High-layers

HLF

LINK/MAC

ANCC

Application

RF interfaces

High-layers

LINK/MAC

User

HLF

APC

QoS

Manager

HLF

APC

ANSC ANCC

HLF

ANSC

Figure 1: Role of the Quality of Service Manager (QoSM)

in the protocol.stack. The QoSM interacts with both ap-

plications and MAC layers of RAT via Application Proﬁle

Commands (APC) and Interface Proﬁle and Status Com-

mands (IPSC), respectively.

All three resilience indicators may be strongly cor-

related: QoS indicators of unstable proﬁles tend to

be difﬁcult to predict; very ﬂexible proﬁles should

exhibit stable QoS indicators that can be accurately

predicted; ﬂexible proﬁles should also be stable, but

stable proﬁles may not be ﬂexible. However, estimat-

ing the inter-dependencies between them is complex

due to the huge cardinality of resources that can de-

ﬁne conﬁguration proﬁles. Additionally, applications

may differ regarding their resilience necessities. For

example, some applications may need a very stable

link with low throughput requirements; in this case

stability and predictability have priority over ﬂexibil-

ity to select an access technology. Conversely, other

applications may require high data rate in all possible

scenarios, so ﬂexibility must have a higher weight to

select the link to be established. Therefore, resilience

indicators are required to be weighted according to

applications needs when the selection process takes

place.

Adaptive Management, the spacial and temporal

feature of resilient systems that allows them to dy-

namically adapt to emergent conditions, reduce un-

certainty, and enhance learning in safe-to-fail manner

(Connelly et al., 2017) is the key to pursue our goal:

ARC applications. The QoSM is in essence the con-

cept of an adaptive management system for proactive

CPS and corresponding conﬁguration of resources in

the available RAT. The following section details the

functionalities and components of the QoSM.

3 QoS MANAGER

FUNCTIONALITIES AND

INTERNAL COMPONENTS

The HLL has been deﬁned to enable uniﬁed process-

ing, offer a uniﬁed interface to the higher layers, and

adapt to the underlying RAT (Zheng et al., 2015a). A

QoS-oriented implementation of this layer can be de-

scribed as a continuous control functionality: receive

QoS requests from upper layers, gather and monitor

QoS performance indicators of available RAT, select

network resources that meet QoS requirements, and

instruct both application and RAT to function coher-

ently with each other. The QoSM proposed in this

position paper is a HLL concept oriented to achieve

ARC safety-critical applications. Details on its tasks,

internal blocks and ﬂow of information are given as

follows.

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

604

3.1 Functionalities

Figure 1 shows the role of the QoSM in an exemplary

service-user connection. High-Layer Frame (HLF)s

are generated by applications at the service side and

go through the QoSM, which determines their re-

quested/speciﬁed QoS, hereinafter referred to as Tar-

get Quality of Service (TQoS), before forwarding

them to low level layers, i.e., LINK, MAC and phys-

ical (PHY). Applications can indicate QoS perfor-

mance requirements, for example, marking Layer-3

header with a code in its Differentiated Services Code

Point (DSCP) ﬁeld as it is done in LTE and 5G archi-

tectures (3rd Generation Partnership Project, 2018).

The QoSM shall be able to read this marking system

and determine QoS requirements in terms of indica-

tors, i.e., throughput, latency, jitter. The QoSM in-

teract with applications via Application Proﬁle Com-

mands (APC). APCs inform applications about the

the achievable QoS, so that they select coherent func-

tional modes matching QoS conditions, e.g., report

the communication throughput to a video streaming

service which shall set video resolution accordingly;

or warn applications of coming degradation of con-

nection quality, so that they can activate safety modes

or mechanisms. Via APC, applications can also spec-

ify QoS requirements (in case they do not use an ex-

plicit marking system in their HLFs), and request sta-

tus of QoS indicators to the QoSM.

Whereas APC enable communication with up-

per layers, Access Network Conﬁguration Com-

mands (ANCC) and Access Network Status Com-

mands (ANSC) are used for the interaction between

the QoSM and MAC layers of RAT. On the one hand

ANSC retrieve link quality, network-related and QoS

indicators directly from each MAC layer of RAT. On

the other hand , ANCC contain the conﬁguration pro-

ﬁle, to be set at each RAT.

As a HLL concept implementation towards ARC

applications for autonomous driving, our proposed

collaborative and distributed QoSM shall:

• Determine the TQoS of applications accessing to

RAT. This is done either by extracting it from

marks in HLFs or directly via APC.

• Receive HLFs from RAT MAC layers and for-

ward them to upper layers.

• Retrieve indicators of link quality, network condi-

tion and QoS from all available RAT via ANSC.

• Quantify and analyze resilience indicators based

on gathered indicators.

• Predict QoS indicators with a given prediction

horizon. The cardinality and complexity of the

variables that determine QoS indicators is ex-

pected to be high: increasing number of RAT

and resources in 5G, numerous interdependent

and parallel applications, highly dynamic environ-

ment and network topology, among others. This

suggests the use of machine learning algorithms

for this purpose.

• Share collected and predicted QoS indicators with

QoSMs at other entities.

• Share the Proﬁle Quality of Experience (PQoE).

This resembles the resilience feature of memory

(Connelly et al., 2017), and indicates the perfor-

mance obtained by setting a giveninterface proﬁle

under speciﬁc conditions, e.g., relative position

and velocity, SNR, noise ﬂoor, current throughput

demand, etc.

• Evaluate the overhead costs of conﬁguration pro-

ﬁles in terms of hardware reconﬁgurationand syn-

chronization times, random network access and

any aspect that may increase communication la-

tency.

• Determine the conﬁguration proﬁle that shall en-

sure the target QoS of running applications with a

very high probability.

• Set conﬁguration proﬁles of available RAT using

ANCC.

• If it is allowed by the application, communicate

with applications via APC to inform QoS condi-

tions.

3.1.1 Functional Components and Interface

Management for V2X Communication

Until now, only generic functionalities of the QoSM

have been described. We now focus on its internal

QoS

Predictor

Profile

Selector

Predicted

QoS

Target

QoS

Interface Manager

Link status

HLF

Interface 1

HetNet

V2X Application

APP

CAM

APP

CAM

HLF

Profile

APC

QoS Manager

Interface 2

Interface N

QoS

Classificator

HLF

Figure 2: Functional blocks of the Quality of Service Man-

ager (QoSM) .

Collaborative and Distributed QoS Monitoring and Prediction: A Heterogeneous Link Layer Concept towards always Resilient V2X

Communication

605

components and their functionalities to achieve ARC

applications in the V2X communication context. Fig-

ure 2 details the functional blocks of the QoSM and

their interactions with both applications and inter-

faces of a HetNet used to transmit V2X data. HLFs

generated at the V2X application side are analyzed

by a QoS classiﬁer in the QoSM to extract the TQoS.

After obtaining the TQoS, the classiﬁer forwards the

analyzed HLFs to the Interface Manager to trans-

mit them over the physical interfaces of the HetNet.

Concurrently, an application is running at every en-

tity generating and receiving periodically Cooperative

Awareness Message (CAM)s, which provide infor-

mation of presence, positions as well as basic status

of communicating Intelligent Transport System (ITS)

stations to neighboring ITS stations that are located

within a single hop distance (European Telecommu-

nications Standards Institute (ETSI), 2019). Using

the information contained in this messages (position,

velocity, acceleration and steer angle) and records of

QoS indicators, the QoS Predictor estimates the per-

formance parameters of the interfaces, i.e., Predicted

QoS, for a given time horizon. Both target and Pre-

dicted QoS are sent to the Proﬁle Selector, which is

the functional block in charge of selecting the conﬁg-

uration proﬁles of RAT.

Figure 3 depicts the concept of conﬁguration pro-

ﬁles, which not only specify technological resources

speciﬁc for each RAT, but also application-speciﬁc

and technology-independentdescription of target per-

formance. Proﬁle parameters may be grouped accord-

ing to speciﬁc requirements of an application. Table 1

exempliﬁes the deﬁnition of proﬁles. Applications re-

quiring high performance are mapped to one of four

proﬁles that aim at obtaining high data rate rather than

very high PDR and low latency. On the other hand, re-

liability and safety proﬁles aim to ensure low PDR at

cost of data rate and latency. The task of the Proﬁle

Selector is to use the predicted QoS to select the pro-

95%

99%

99.5% 99.9%

Max

10Gbps

1Gbps

0.5Gbps

20ms

10ms

100ms

1ms

Value 3

Value 1

Value 2

Value 4

Configuration

Profile

PDR

Throughput/Data rate

Latency

Parameter N

traffic

steering

dual-

connectivity

FEC Interleaving

technological resources

Figure 3: Concept of Conﬁguration Proﬁle. Target QoS in-

dicators and technological resources are grouped to deﬁne

a more abstract representation.

ﬁle that ensure the Target QoS with very high proba-

bility, and inform the application via APC whether the

required performance can or cannot be met to adjust

its requirements accordingly.

Conﬁguration proﬁles are agnostic descriptions,

i.e., technology-independent. However, RAT of the

HetVNET have to be conﬁgured using their own

set of speciﬁc commands. The task of the Inter-

face Manager is to translate the high-level descrip-

tion and parameters of proﬁles into interface-speciﬁc

settings and corresponding conﬁguration commands.

Then, agnostic commands coming from the pro-

ﬁle selector, e.g., ”SetPerformanceProﬁle”, must be

mapped to speciﬁc technology commands such as

”SetWiFiChannel(5gHz), SetWiFiOFDM(), SetWiFi-

TransmitPower(13.5dBm)”. The tasks of the Inter-

face Manager include routing HLFs via the selected

interface(s) and retrieving measurements, link status

and current conﬁguration settings of the RAT.

QoS Predictor and Proﬁle Selector are the core el-

ements of the QoSM. The capacity of the QoS Predic-

tor to accurately estimate QoS indicators, determines

the timely selection of suitable conﬁguration proﬁles

performed by the Proﬁle Selector. On the other hand,

the Proﬁle Selector must choose ”wisely” among a set

of conﬁguration proﬁles, and shall be able to mod-

ify them and add new entries to the proﬁle table. To

achieve these goals, we propose a collaborative and

distributed scheme for QoSM functionalities. In this

approach, predicted QoS, PQoE and link status mea-

surements are multi-casted along a group of QoSMs.

This provides prediction and proﬁle selection func-

tions with more data to accomplish their tasks. The

following session explains how such scheme shall be

developed and how it would contribute to achieve re-

silient V2X communication.

Table 1: Proﬁle deﬁnition. Example of performance pa-

rameters that are combined to deﬁne high-level descrip-

tions based on application requirements such as high per-

formance, reliability and safety.

Application Data PDR latency (ms)

Requirement Rate

High Maximize 95 5

Performance 8 Gbps 95 10

6 Gbps 95 10

4 Gbps 95 10

Reliability 2 Gbps 99 20

1 Gbps 99 20

100 Mbps 99.5 50

Safety 10 Mbps 99.9 50

1 Mbps 99.99 100

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

606

4 COLLABORATIVE AND

DISTRIBUTED MONITORING,

PREDICTION AND PROFILE

SELECTION

Figure 4 illustrates a group of 5 vehicles that can

communicate with each other via two links: over

Vehicle-To-Vehicle (V2V) communication, e.g., WiFi

802.11p, and the infrastructure using mobile technol-

ogy, e.g., acLTE. Red and blue arrows indicate active

links. Now, let us assume the following hypothetical

scenario:

• Car B is located at 10 m and 45

◦

from car A.

• Car C is located at 13 m and 57

◦

from car D.

• Car B is located at 16 m and 72

◦

from car E. They

are currently experiencing very good performance

while implementing the ﬁrst conﬁguration proﬁle

in Table 1.

According to current positions, velocities and ac-

celerations of vehicles informed via CAMs, commu-

nicating pairs (A,B) and (D,C) are expected to be lo-

cated in 100 ms at the same current relative position

of pair (E,B), i.e., 16 m and 72

◦

from each other. If

cars A, B, C and D are experiencing increment of the

performance demanded by their applications, it would

be very useful for their QoSMs to receive the cur-

rent PQoE of pairs (E,B), so that their proﬁle selectors

can conﬁgure timely their RAT to fulﬁll the potential

future conditions. The same reasoning can be used

for the Vehicle-To-Infrastructure (V2I) communica-

tion of vehicles A and C, and also in the prediction

of QoS performance at the QoS predictor. The gen-

Figure 4: Collaborative and distributed scheme for monitor-

ing, prediction and proﬁle selection. Enabling task distribu-

tion and data sharing of data useful for QoS prediction and

proﬁle selection, help to obtain resilient V2X communica-

tion.

Figure 5: Scaling collaborative and distributed scheme.

Shared data of V2X and V2I communication can be trans-

mitted beyond local coverage. Access network entities such

enhanced Node B (eNB)can beneﬁt from shared data to per-

form resource allocation and reserve resources for groups of

cars approaching to its coverage range.

eral reasoning is that increasing the amount of useful

data should derive in a more accurate predictions and

resilient communication.

The problem of grouping entities can be addressed

by analyzing the correlation of conditions, predic-

tions and proﬁles: entities with similar QoS indica-

tors, conﬁgured proﬁles, performance demand and

mobility should be encourage to collaborate. Addi-

tionally, some entities may be able to produce more

accurate predictions due to their relative positions,

number of links or time of membership within the

group. Then, each group could select the best en-

tity to rule group’s proﬁle conﬁgurations and be rep-

resentative to share group’s PQoE, QoS prediction

and link status measurements with other groups. This

would reduce the trafﬁc of shared data and add fea-

tures of self-organization to HetVNETs. Cooperative

and non-cooperativegame theory algorithms could be

considered for the selection of groups’ members.

4.1 Scaling the Collaborative and

Distributed Scheme beyond Local

Coverage

Whereas the information contained in CAMs is lim-

ited to be received by a geographically-deﬁned group

of entities (European Telecommunications Standards

Institute (ETSI), 2019), data such PQoE, link status

measurements and predicted QoS may be useful for

groups of entities located out of the vicinity. In mo-

bile networks such as LTE, an evolved Node B (eNB)

Collaborative and Distributed QoS Monitoring and Prediction: A Heterogeneous Link Layer Concept towards always Resilient V2X

Communication

607

providing mobile service to a group of vehicles can

share equivalent information with other nodes regard-

ing individual or global resource allocation. This is,

the current radio resources being demanded by the

served group and a prediction of the potential future

demand could be reported to other eNBs on the rode.

Hence, eNBs that will soon serve the same group

could reserve the same or equivalent resources in ad-

vance, providing lower hand-over times or the pos-

sibility to instruct vehicles to start safety modes in

case resources cannot be allocated. Figure 5 shows

two groups of communicating cars whose V2I com-

munication is being served by two different eNBs.

The backhaul network communicating them can be

used to transmit collaborative data between nodes or

for QoSM at vehicles. Such potential implementation

shows how scalable our collaborative and distributed

scheme can be.

5 CONCLUSIONS AND FUTURE

WORK

In this position paper,we propose a conceptfor a HLL

implementation. The proposed scheme, called Qual-

ity of Service Manager (QoSM), aims at obtaining re-

silient applications for V2X communication. We ar-

gue that using QoS indicators to quantify and eval-

uate resilience, and adopting the concept of Always

Resiliently Connected will lead to resilient safety-

critical applications. To achieve this, our approach

proposes a collaborative and distributed solution that

seeks to increase and diversify the sources of informa-

tion for prediction of QoS parameters and selection of

conﬁguration settings of HetVNETs. Our approach is

highly scalable since it enables QoSMs to group hi-

erarchically, as well as the deﬁnition of technology-

independent proﬁles, which results in agnostic man-

agement of HetVNET.

Future work includes the implementation and re-

sults analysis of the proposed scheme to ensure the

communication performance for collective function-

alities of autonomous driving such as sensor-data

sharing and see-through application.

REFERENCES

3rd Generation Partnership Project (2018). 3GPP TS 23.501

(V15.2.0): System Architecture for the 5G System

(Release 15). Technical Speciﬁcation.

Connelly, E. B., Allen, C. R., Hatﬁeld, K., Palma-Oliveira,

J. M., Woods, D. D., and Linkov, I. (2017). Features

of resilience. Environment Systems and Decisions,

37(1):46–50.

European Telecommunications Standards Institute (ETSI)

(2019). ETSI EN 302 637-2 (V1.4.1): Intelligent

Transport Systems (ITS); Vehicular Communications;

Basic Set of Applications; Part 2: Speciﬁcation of Co-

operative Awareness Basic Service.

Fraunhofer ESK Institute (2018). Safe autonomous

systems–resilient systems. https://www.esk.

fraunhofer.de/en/research/safe-autonomous-systems-

resilient-systems.html.

Gustafsson, E. and Jonsson, A. (2003). Always best con-

nected. IEEE Wireless Communications, 10(1):49–55.

Lahby, M., Baghla, S., and Sekkaki, A. (2015). Survey and

comparison of MADM methods for network selection

access in heterogeneous network. In in Proccedings

7th International Conference on New Technologies,

Mobility and Security (NTMS).

Linkov, I. (2018). Building and Quantifying Resilience in

Complex Systems. 6th OECD World Forum on Statis-

tics, Knowledge and Policy.

Louta, M. and Bellavista, P. (2013). Bringing always

best connectivity vision a step closer: challenges

and perspectives. IEEE Communications Magazine,

51(2):158–166.

Louta, M., Zournatzis, P., Kraounakis, S., Sarigiannidis, P.,

and Demetropoulos, I. (2011). Towards realization of

the ABC vision: A comparative survey of Access Net-

work Selection. In in Proccedings IEEE Symposium

on Computers and Communications (ISCC).

Mohamed, L., Leghris, C., and Abdellah, A. (2012). A Sur-

vey and Comparison Study on Weighting Algorithms

for Access Network Selection. In in Proccedings 9th

Annual Conference on Wireless On-Demand Network

Systems and Services (WONS).

Roy, A., Chaporkar, P., and Karandikar, A. (2018). Optimal

radio access technology selection algorithm for lte-

wiﬁ network. IEEE Transactions on Vehicular Tech-

nology, 67(7):6446–6460.

Trestian, R., Ormond, O., and Muntean, G.-M. (2012).

Game Theory-Based Network Selection: Solutions

and Challenges. IEEE Communications Surveys & Tu-

torials , 14(4):1212 – 1231.

Wang, M., Chen, J., Aryafar, E., and Chiang, M. (2016).

A Survey of Client-Controlled HetNets for 5G. IEEE

Access, 5.

Zheng, K., Member, S., Zheng, Q., Chatzimisios, P., Xiang,

W., and Zhou, Y. (2015a). Heterogeneous Vehicular

Networking: A Survey on Architecture, Challenges,

and Solutions. IEEE Communications Surveys & Tu-

torials, 17(4):2377–2396.

Zheng, K., Zheng, Q., Yang, H., Zhao, L., Hou, L., and

Chatzimisios, P. (2015b). Reliable and efﬁcient au-

tonomous driving: the need for heterogeneous ve-

hicular networks. IEEE Communications Magazine,

53(12):72–79.

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

608