F R A U N H O F E R - G E S E L L S C H AF T Z UR F Ö R D ER U N G D E R AN G E W A N D T E N F O R S C H U NG E . V .

ON THE ROAD TO 6G:

DRIVERS, CHALLENGES AND
ENABLING TECHNOLOGIES
A Fraunhofer 6G white paper

The work presented is part of the Fraunhofer lighthouse project 6G SENTINEL: Six-G
Enablers: Flexible Networks, THz Technology and Integration, Non-Terrestrial Networks,
SidElink, and Localization. Additional information on the status of the project is available
at: www.iis.fraunhofer.de/6g-sentinel

Fraunhofer Institute for Integrated Circuits IIS

Am Wolfsmantel 33
91058 Erlangen, Germany

Contact
Bernhard Niemann
communicationsystems@iis.fraunhofer.de

www.iis.fraunhofer.de/6g-sentinel
Contributors (in alphabetical order)

Corici, Marius-Iulian Fraunhofer FOKUS

Franke, Norbert Fraunhofer IIS
Heyn, Thomas Fraunhofer IIS
Kontes, Georgios Fraunhofer IIS
Leyh, Martin Fraunhofer IIS
Magedanz, Thomas Fraunhofer FOKUS
Maaß, Uwe Fraunhofer IZM
Mikulla, Michael Fraunhofer IAF
Niemann, Bernhard Fraunhofer IIS
Peter, Michael Fraunhofer HHI
Roth-Mandutz, Elke Fraunhofer IIS
Schubert, Colja Fraunhofer HHI
Yammine, George Fraunhofer IIS

Version 1.0 November 2021 2 | 15

On the road to 6G

Deployment of 5G is still on its way while discussion about 6G technologies and

corresponding research activities have started all over the world. Now is the time to work
on the enabling technologies and fundamental building blocks of the new mobile
communication standard and to join the worldwide discussion on use cases,
functionalities and key-performance indicators (KPIs) of 6G. The large interest of
academia and industry alike can be seen from the substantial amount of white papers
and publications [2]-[13].

Standardization in 3GPP is just about to enter the second phase of 5G with 5G Advanced,
starting from Release 18 beginning of 2022. 6G related activities are assumed to start
around 2025, while first 6G deployments can be expected for 2030. The evolution of
wireless mobile communication standards from 1G to 6G is depicted in Fig. 1. While 1G
and 2G were focused on speech, data services and internet access were added in 3G.
4G brought the rise of the mobile internet whereas 5G is focusing on machine-to-
machine communication and the internet of things. With 6G it is expected that humans
and their needs will be in the center, once again (see section “New perspectives on 6G
use cases”).

Fig. 1: Evolution of wireless

mobile communication
standards

An important trend in 6G is the convergence of the physical and the human world with
the digital domain. Frequently discussed use cases for 6G are digital twins, virtual and
augmented reality (XR), tele-presence, tele-operation and autonomous driving. To enable
these use cases, a significant increase of 10x or more in the traditional KPIs like peak data
rate, reliability, latency, connection density, and localization accuracy is required from 5G
to 6G. On the technological side, the following enablers are crucial to achieve the
required improvement in the KPIs:

 Terahertz radio technologies for ultra-high data rates

 Flexible networks for situationally-adapted network availability
 Fully integrated localization utilizing reinforcement learning for greater accuracy
 Optimized network architecture for best service quality and reliability

These enabling technologies are the foundation needed to respond to the challenges
and requirements of 6G and, therefore, will have to be developed in a timely manner to
be ready for integration as part of standardization as well as later products. If 6G
technology can deliver its promises, it will stimulate and open new application fields and
businesses.

Version 1.0 November 2021 3 | 15

Fig. 2 highlights the approach for bridging the gap between 5G and 6G with high
innovation. Up to 2025, when standardization of 6G is expected to be ready to start, a
very high potential for innovation is present in different domains and directions. In order
to innovate in such a pre-standardization environment it is utmost important to consider
the requirements of the network of 2030 and to break out of the specific 5G technology
thinking. This step can be achieved only by conceptual development and practical
feasibility assessments in a highly open ecosystem. Specifically, a new framework has to
be considered where the basic assumptions of communications are highly different from
5G.

Fig. 2: Bridging the gap

between 5G and 6G
standardization with high
innovation

An intense collaboration and exchange of concepts and designs between the different
domains represents a way in which a coherent 6G network development can be
accelerated. Only a holistic development including access, core, transport, devices and
application domains could produce the expected results.

The Fraunhofer lighthouse project 6G SENTINEL [1] is addressing the specific technical
challenges in these directions within a pragmatic, results oriented approach, aiming to
boost the research activities, to prepare the standardization foreseen to start in 2025 and
to accelerate the development and the foreseen adoption of commercial 6G networks in
the year 2030.

What this white paper contains

 New perspectives on 6G use cases
 A high-level classification of challenges and requirements
 An approach and a roadmap for 6G R&D activities
 First steps towards a 6G architecture
 An applied research best-practice R&D roadmap
 The status and achievements of the 6G SENTINEL project

Version 1.0 November 2021 4 | 15

New perspectives on 6G use cases
While being still in an early phase regarding the potential use cases which will make 6G
technology successful, due to the currently widening deployment of 5G networks, a set
of highly potential directions could be identified for which the current standard
commercial networks are not able to satisfy their requirements.

The classification of use cases presented here aims to entice the readers to change their
perspective on how networks will transform. This does not exclude the further evolution
of the 5G use cases which will happen in parallel, instead it potentiates it with a longer
perspective addressing the communication needs of the year 2030.

Starting from the 4G Machine Type Communication (MTC) and 5G Ultra Reliable Low-
Latency Communication (URLLC), a strong diversification of classes of services was
observed beyond only increasing the network capacity, requiring specific network
characteristics which could not be immediately fulfilled with the same infrastructure.
While still expecting to serve more devices, using higher network capacity with more
reliability or lower delay, the next generation of applications will require a mixture of
these capabilities. For example, new enhanced Mobile Broadband Services (eMBB) such
as Augmented Reality (AR) / eXtended Reality (XR) would also require a high level of
reliability and a fairly low delay in addition to the network capacity. Similarly, the massive
MTC would have to extend itself towards reliability and low delay to be able to handle
fine-granular automatic environment management and to increase the network capacity
when going in the direction of autonomous mobility and video-based insight generation.

Fig. 3: Classification of 6G use

cases

However, the key driver for 6G will remain the extended network capacity. And this can
be achieved only through new, Terahertz (THz) spectrum-based technologies. Be it
capacities of 4 Tbit/s for AR/XR, the under 100 µs delay for industrial or holographic
presence, a 7-nines reliability, or < 1 cm-precision localization. For all of these use cases,
no matter how much the compute capacity of the terminals will rise in the next years,
an even higher compute capacity will be needed from the network. This continuous
communication and the adoption of an increased number of devices will require a
significant enhancement to the current 5G NR networks.

A key feature foreseen in 6G is the tight integration of localization in the network. Not
only will position information be used for the end-user location services, this information
can be exploited as well to improve the overall performance of the network, where early

Version 1.0 November 2021 5 | 15

coarse information in one domain can be used to improve the other (e.g., coarse location
information to enable beamforming with less feedback from the mobile device).
Moreover, location information can be aggregated and, with the help of machine
learning algorithms, used to coordinate the network. This gives rise to a “joint
communication and positioning” concept.

With 5G, a first step was made towards the support of private and non-public networks.
This opened the possibility to cover on-premise type of deployments resulting in more
localized networks and in the option to develop full campus or even regional networks.
However, this also opened the option to cover additional mobile use cases. For 6G, the
next step is to be able to communicate with other on-premise locations, to establish
larger scale networks or global networks, resulting in an ultra-flat deployment where the
coverage is tethered to specific areas where the communication is needed (e.g., factory
campuses around the world), integrated within a single network by using backhaul
connections (similar to current large enterprise networks).

Another important aspect of 6G will be ubiquitous availability and to extend the coverage
to reach a more global connectivity. 3GPP already started with the integration of satellites
in the 5G system (called Non-Terrestrial Networks, NTN). This trend will continue with
including and interconnecting multiple different satellite orbits and the network
architecture will be evolved towards 6G, where also various airborne (UAVs, planes,
HAPs) and spaceborne platforms (different types of satellites) will be interconnected with
the terrestrial 6G network on a multi-layer (3D) architecture, requiring a dynamic, flexible
RAN and core network architecture.

This range of new small networks span from fixed deployments such as currently
foreseen for factory shop floors, temporary networks with additional base stations,
increasing the local capacity, nomadic deployments, providing coverage for a given
interval within a location such as a music festival or construction site, up to mobile
deployments as needed by Public Protection and Disaster Relief (PPDR), public transport,
logistics and maritime.

The envisioned future network flexibility along with the contrastive requirements of 6G
services and applications presuppose a complex interplay between large-scale
ecosystems of software and hardware network components, rendering classical
theoretical approaches unable to seamlessly scale to the massive problem size. Artificial
intelligence can alleviate this problem by being a key enabler towards self-adapting, data-
driven and scalable orchestration of heterogeneous network services, resources and
applications.

Fig. 4: 6G network
deployments

Version 1.0 November 2021 6 | 15

Challenges and enabling technologies
Like previous generations, 6G is promising to fundamentally change how the consumers
and businesses communicate, paving the way for a next generation of use cases able to
capitalize on the speed, capacity, latency, and flexibility it offers as well as on the parallel
compute, visualization, and artificial intelligence developments. It is expected that only a
mix of heterogeneous technologies integrated within a comprehensive system can
properly respond to the requirements.

A first summary of these technologies is presented here aiming to target the research in
the expected direction.

Table 1: 6G requirements and

Technical Requirements Enabling Technologies enabling technologies
Challenge
Higher Data Cover additional, higher Decreasing the reception range,
Rates frequencies (mmW, THz) physical impairments, smaller cells
Schedule available resources Granular cell-free resource
scheduling
Ubiquitous Larger Cell Size Converging satellite and airborne
Availability
Smaller nomadic cells Dynamic spectrum allocation
Support for split core networks
Integrated backhaul management
Advanced Positioning and sensing Tight integration of sensing,
cognitive insight positioning, and AI
Fluid data exchanges Secure data acquisition and exchange
Data storage and curation
Integration of AI in network Automatic, self-learning end-to-end
management decision chains for robust
performance and security
management.
Integrated RAN Integrating RAN and core mobility
management management and QoS
Robust Flexible, organically adapted Using software-only paradigms to
communication core network functionality design the core network service
End-to-end coherent service Integrating backhaul technologies as
delivery part of the core network
Ultra-reliable core services Redesigning the core network for
native reliability
Clear split of state Adapted mechanisms for state
information and stateless sharing and state migration
components
Heterogeneous Security and trust Redefinition of the interaction
infrastructure between infrastructure and services
owners
Deterministic QoS Extending backhauls with
deterministic capabilities

Version 1.0 November 2021 7 | 15

 Multi-governance Coherent system governance in
multi-administrated environments
Sustainability Requires reduction of Network and UE energy reduction
carbon footprint of mobile and power optimization
networks
Ultra-adaptable networks Reducing the network overhead by
very fast resource allocation and
adaptation decisions.
Ultra-precise Finer temporal resolution Higher frequencies (THz) allows the
positioning usage of signals with large enough
bandwidth to provide the temporal
resolution
Finer angular resolution THz signals enable ultra-massive
MIMO antenna configurations which
provide a very-fine angular resolution
Support for very Denser networks Distributed and cell-free massive
large number of MIMO networks
devices

In the following, a brief introduction to three important technological enablers, namely

Terahertz technologies, flexible networks and integrated localization along with the
associated research challenges will be given. The fourth research area, 6G network
architecture, will be discussed in greater detail in the next section “Towards a 6G
network architecture”.

Terahertz technologies for ultra-high data rates

To achieve the envisioned data rates of up to several Tbit/s in wireless links, bandwidths
of 10 GHz and more are required. This can only be achieved at frequencies above
100 GHz, commonly referred to as Terahertz. The transmission channel for frequencies
beyond 100 GHz, especially for indoor environments and sidelink communication
between vehicles in the presence of mobility, is not sufficiently researched and new
channel models need to be developed. Furthermore, the high frequencies increase the
Doppler-rate while the large bandwidth leads to hardware impairments like phase noise
and non-linearities that all need to be compensated. Novel approaches for integrated
THz-transmitters with sufficient transmit-power like, e.g., a combination of GaAs and
GaN or photonic InP technologies need to be investigated. In addition, cost efficient
packaging technologies are required to decrease the cost of THz-modules. ([14], [15])

Flexible networks for unlimited 6G availability

The combination of the demand for heterogeneous infrastructure and ubiquitous

coverage requires a flexible network with tight integration of a Non-Terrestrial Network
subsystem.

In 6G, a more seamless integration of the different network nodes is expected than in
5G, where only the first steps for integrating of new non-terrestrial elements is
standardized. With this level of integration, a truly global connectivity can be realized.
Generally, three different layers of network elements (in terms of altitude) can be
assumed: Ground-based platforms, airborne platforms, and spaceborne platforms. The

Version 1.0 November 2021 8 | 15

complexity of the network will introduce different challenges, which need to be analyzed
and solved to be able to deploy such 6G systems. On the system level, it is necessary to
be able to add and remove new network nodes on a dynamic basis (e.g., LEO satellites,
UAVs). Also, the inter-satellite routing between all these platforms will be a challenge.
The delay between the platforms introduced by their motion will vary and the need for
an accurate position of all elements is mandatory for proper resource and interference
management in coexistence and spectrum sharing scenarios with terrestrial and non-
terrestrial nodes.

Starting from the 5G core network, innovative concepts in the directions of secure end-
to-end communication, trust in the infrastructure, reliable and deterministic backhauls,
subscriber state distribution and data layer need to be developed. These features come
to complete the 5G architecture and provide a basis for the 6G one towards a smooth
end-to-end software network deployment on distributed network infrastructures.

Fully integrated localization in 6G networks for greater accuracy

Positioning in 6G networks is mainly driven by three technology enablers, namely

distributed massive MIMO, jointly-processed coordinated multipoint transmission and
ultra-dense networks which form the basis of cell-free massive MIMO systems. Cell-free
massive MIMO does not entail a fixed network, but dynamically-changing clusters of
access points that follow the user. With this paradigm shift, new opportunities and
challenges rise. Fixed base stations that could be used as anchors for positioning are now
dynamically changing. New fitting procedures, protocols, and functions need to be
developed. Along with the tighter integration of localization in 6G networks comes also
higher signaling overhead for the measurement of positioning-related signals. Novel
approaches for a common reference signal for communication and positioning are
required. Terahertz signals with very-large bandwidth and ultra-dense networks offer the
possibility of ultra-precise positioning (in comparison, cm-precision localization is feasible
with millimeter-wave signals [16]). Research on suitable ultra-massive MIMO antenna
arrays, their characteristics and their impact on positioning performance is needed.
Furthermore, AI methods, especially reinforcement learning (RL) for beam-management,
are promising approaches for performance optimization. Owing their plug-and-play
nature, RL-based methods are able to solve complex dynamical problems with quality-
of-service and safety constraints without the need of hard-to-obtain models of the
system. Moreover, they offer the possibility to extract interpretable decision rules [17].

Version 1.0 November 2021 9 | 15

Towards a 6G network architecture
A 6G architecture must consider how to support the selected use cases within the
infrastructure context of 2030 when large amounts of compute resources would be
available at different locations spanning from mobile phones, campus network back-
ends, satellite payloads or on-demand in different mobile vehicles. For this, the network
will have to morph to the existing compute and networking resources and to migrate
specific subscriber contexts [2], [3].

Fundamental trends for the network of 2030

 a continuous decrease in compute costs and increase in capabilities
enabling the virtualization of even more components
 a further increase in the software development paradigms enabling a 10-
times faster software development
 extreme increase of the access and backhaul radio capabilities resulting in
a significant larger set of options to be deployed
 large scale deployment of low orbit satellite networks, changing the
perspective on global networks with ubiquitous availability

Fig. 5: Flexible networks high

level functionality

Furthermore, it must include the current 5G functionality supporting the connectivity

service such as authentication and authorization, connection and mobility management
or session management. To be able to respond to the services’ diversification, the
architecture must maintain an openness to add new services in the form of new network
functions. Albeit the 5G architecture aimed at this functionality with the adoption of a
Service Based Architecture (SBA), such an open architecture is not possible as the new
services are highly dependent on the offering of the existing ones. Ultimately, any
extension requires the upgrading of existing services too, which leads to an extensive
standardization process.

Version 1.0 November 2021 10 | 15

 Table 2: Requirements and
Requirements Challenges challenges of 6G network
architecture
Infrastructure-free  Deployment on any hardware
implementation  Transparent network bindings
 Account for noticeable service characteristics
Complexity reduction  Re-grouping of functionality in new elements
 Remove functionality replicas
 Reduce the number of exchanged messages per
procedure
 Uniform support of RAN and Core control in the edge
 Service Based Architecture to the UE
Organic growth  Regrouping of the functionality
 Split processing and state elements

Simplify the addition of  Reduce functional dependencies

new functionality  Reduce the number of interfaces

Very fast scaling  Native support for load balancing

 Effective state sharing
High parallelization  Easy to split load across multiple worker entities
capacity  Able to parallelize across distributed infrastructures

Continuous integration /  Graceful deployment of functionality during the

Continuous deployment runtime of the system – on ground, in the air, in the
sky
Network management  Same type of components, uniform policies for
simplification scaling, uniform configurations, easy to adapt to
automation
Support the existing 5G  Access control, authentication, and authorization
functionality  Connection management
 Mobility management
 Session management
 RAN capabilities management
 Data path forwarding
 Lawful interception and charging

To be able to address these requirements in a graceful manner the overall architecture

must be re-thought in the context of the parallel software engineering developments.
This would presume to analyze from a conceptual level the concept of network functions.

As the name says, network functions are functionally defined based on their input and
output interfaces, the transfer function which processes the messages and the subscriber
state enabling the appropriate processing. Ultimately, a network function is a system by
itself which works independently of the other network functions. As they have
standardized input and output interfaces, they can be implemented and tested in
isolation from other network functions enabling easy interoperability tests.

However, the complete independence of each network function is introducing many

limitations. Maintaining independent state in each of the components is creating an
information multiplication as well as it requires many messages to be exchanged
between the components. Furthermore, the large number of messages create additional

Version 1.0 November 2021 11 | 15

bindings between components, so instead of a highly flexible system we in fact have a
very monolithic one with no possibility to deploy new services without cascading
modification effects through the system.

Fraunhofer lighthouse project 6G SENTINEL

Based on the key drivers, challenges and enabling technologies identified before, new
answers are needed to master the rapidly increasing performance requirements. 6G
SENTINEL aims not only to push existing 5G technologies further, but also to develop
brand new approaches that will help applications such as virtual reality, digital twins and
autonomous driving achieve a genuine breakthrough with 6G.

In the Fraunhofer lighthouse project 6G SENTINEL (Six-G Enablers. Flexible Networks, THz
Technology and Integration, Non-Terrestrial Networks, SidElink, and Localization), the
five participating Fraunhofer Institutes are developing key technologies for the upcoming
6G mobile communications standard.

6G SENTINEL is addressing the most urgent challenges of mobile communications with

practical routes to effective solutions (Fig. 6).

Fig. 6: 6G SENTINEL: challenges

and enabling technologies
from 5G to 6G

6G SENTINEL is targeting improvements to device antennas and front-end modules. It

also seeks to optimize transmission technologies in the dynamic, heterogeneous
deployments of the radio access network (RAN) and increase the flexibility of the core
network. This means the project will support further developments of all relevant
components of a mobile communications network.

The Fraunhofer institutes bring to the project complementary expertise from the fields
of terrestrial and satellite radio access networks, localization, core networks,
semiconductor technologies for THz communication and electronics packaging. This
creates a unique conjunction of application know-how, technology expertise for
individual aspects of a 6G network and overall systems expertise.

Version 1.0 November 2021 12 | 15

 Fig. 7: 6G SENTINEL:
technology expertise for 6G

6G SENTINEL follows a very pragmatic and agile approach towards technology

development stemming from the large experience in working together with the industry
and within standardization from Fraunhofer. Alongside, 6G SENTINEL aims to provide a
first robust perspective on how 6G will evolve as well as to act as a lighthouse to the
practical 6G developments. A very agile roadmap was put in place with significant results
during 2021 and 2022 in terms of conceptual developments and a first demonstration
of the new developments in September 2022. With this, 6G SENTINEL is acting as
pioneering R&D within the applied research community.

Fig. 8: 6G SENTINEL roadmap

and milestones

Version 1.0 November 2021 13 | 15

 What 6G SENTINEL provides
 Independent assessment of the 6G worldwide developments
 Design and specification of functionality addressing the foreseen
infrastructure of 2030
 Integrated developments across heterogeneous technologies by world-
recognized specialists
 Pragmatic technology development in the following areas:
 THz communications
 flexible networks
 Practical demonstrations of the technologies, showcasing how
technologies work beyond only concepts and use cases lists
 Open environment to exchange ideas and to enable a common
development of 6G
 Acceleration of the 6G environment development through supplying a
comprehensive environment in which own research can be embedded

Version 1.0 November 2021 14 | 15

Further reading
[1] 6G SENTINEL Webpage, https://www.iis.fraunhofer.de/6g-sentinel
[2] E. Bertin, N. Crespi, T. Magedanz (editors), “Shaping Future 6G Networks: Needs,
Impacts and Technologies”, ISBN: 978-1-119-76551-6, Nov 2021, Wiley-IEEE
Press, https://www.wiley.com/en-
be/Shaping+Future+6G+Networks%3A+Needs%2C+Impacts%2C+and+Technolo
gies-p-9781119765516
[3] Corici, M. et al., “An Ultra-Flexible Software Architecture Concept for 6G Core
Networks”, in Proceedings of IEEE 4th 5G World Forum (5GWF) 2021, pp. 1-5,
IEEE, October 2021.
[4] ITU Focus Group NET-2030, “A Blueprint of Technology, Applications and Market
Drivers Towards the Year 2030 and Beyond”, https://www.itu.int/en/ITU-
T/focusgroups/net2030/Documents/White_Paper.pdf
[5] NGMN Alliance, “6G Drivers and Vision”, V1.0, https://www.ngmn.org/wp-
content/uploads/NGMN-6G-Drivers-and-Vision-V1.0_final.pdf
[6] 5G Infrastructure Association (5G IA), “European Vision for the 6G Network
Ecosystem”, V1.0, 2021-06-07, https://5g-ppp.eu/wp-
content/uploads/2021/06/WhitePaper-6G-Europe.pdf
[7] Hexa-X Deliverables, https://hexa-x.eu/deliverables/
[8] 6G Flagship (University of Oulu) White Papers, https://www.6gchannel.com/6g-
white-papers/
[9] Samsung, “6G The Next Hyper-Connected Experience for All”,
https://cdn.codeground.org/nsr/downloads/researchareas/20201201_6G_Vision_w
eb.pdf
[10] NTT Docomo White Paper, “5G Evolution and 6G”, Feb 2021 (Version 3.0),
https://www.nttdocomo.co.jp/english/binary/pdf/corporate/technology/whitepaper
_6g/DOCOMO_6G_White_PaperEN_v3.0.pdf
[11] 5G Americas,”Mobile Communications Towards 2030“, White Paper, November
2021, https://www.5gamericas.org/wp-content/uploads/2021/11/Mob-Comm-
Towards-2030-WP.pdf
[12] ATIS Next G Alliance, Action paper ”Promoting U.S. Leadership on the Path to
6G“, May 2020, https://www.atis.org/wp-content/uploads/2020/07/Promoting-
US-Leadership-on-Path-to-6G.pdf
[13] Rohde & Schwarz (Dr. Nishith D. Tripathi, Dr. Jeffrey H. Reed), “5G evolution – on
the path to 6G”, Whitepaper Version 01.00
[14] C. Castro, R. Elschner, T. Merkle, C. Schubert and R. Freund, “Experimental
Demonstrations of High-Capacity THz-Wireless Transmission Systems for Beyond
5G”, in IEEE Communications Magazine, vol. 58, no. 11, pp. 41-47, November
2020, doi: 10.1109/MCOM.001.2000306.
[15] A. -A. A. Boulogeorgos et al., ”Terahertz Technologies to Deliver Optical Network
Quality of Experience in Wireless Systems Beyond 5G“, in IEEE Communications
Magazine, vol. 56, no. 6, pp. 144-151, June 2018, doi:
10.1109/MCOM.2018.1700890.
[16] G. Yammine, M. Alawieh, G. Ilin, M. Momani, M. Elkhouly, P. Karbownik, N.
Franke and E. Eberlein, “Experimental Investigation of 5G Positioning Performance
Using a mmWave Measurement Setup”, in 11th International Conference on
Indoor Positioning and Indoor Navigation, Spain, November 2021.
[17] L. Schmidt, G. Kontes, A. Plinge, C. Mutschler, “Can You Trust Your Autonomous
Car? Interpretable and Verifiably Safe Reinforcement Learning”, in Proceedings of
the 2021 IEEE Intelligent Vehicles Symposium (IV), Japan, July 2021.

Version 1.0 November 2021 15 | 15