Presentations given at the
Workshop 6: European and Taiwanese Cooperation on 5G
Wednesday, 19 June 2019, at EUCNC 2019 in Valencia, Spain.
All presentations atre available.
4. Standard Timeline
3
IMT-2020
(2015.6)
Official 5G
Approved
2020
2018 2019 2020 2021 2020
Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
R17 Work Scope discussions
R-15
R-15 late
drop ASN.1
R-16 freeze R-17
freeze
R-17 freeze
ASN.1
R-16 freeze
ASN.1
R-15 late
drop freeze R-17 RAN1
freeze
R-16 RAN1
freeze
5. 5G Commercial Launch Plan
4
Pre-commercial
2017 2018 2019 2020
Commercial
Rel-15 (5G phase I) Rel-16 (5G phase II)
mobile
CMCC : 5 cities 500 sites
CU : 16 cities 600 sites
Oct. FWA
April mobile launch
Trial
Dec. CPE
Time : 2018H2
Freq. : 28G/39GHz
Time : 2018 Dec.
Freq. : 3.5G/28GHz
Time : 2020
Freq. : Sub 6G,
28GHz
Time : 2019
Freq : Sub 6G,
26GHz
Time : 2020
Freq. :
2.6/3.5/4.8GHz
AT&T 19 cities
T-Mobile : 30 cities
Sprint : 6 citiesVerizon : 4 cities
Non-standalone standalone Phase II complete
Source:5GO collected
6. Global 5G Spectrum – Sub 6GHz
• Consider 5G single carrier bandwidth will be >= 100MHz in Sub 6GHz, at
least 300MHz is need for market innovation and competition.
5
Considered for 5G by global regulators
300MHz
300MHz
600MHz 500MHz
400MHz
500MHz
China
USA
Japan
Korea
EU
3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5.0 GHz
450MHz
Taiwan 500MHz 400MHz 500MHz
Further Studied by MOTC, Taiwan
150MHz
CBRS (LTE or 5G)
ITU
Global primary MS band
Non-global primary MS band
Source : 5GIA, FCC, IMT-2020(PG), 5GMG, 5G Forum, May 2017. and MOTC Taiwan, June 2017.
Edit by 5GO and MIC.
150MHz
TD-LTE
120MHz
3.3-3.4 indoor use
MOEA 5G R&D Spectrum (3.4-3.6GHz)
200MHz
100
MHz
7. Global 5G Spectrum – Above 6GHz
6
3 GHz
3.25 GHz 6.5 GHz
1.6 GHz 3 GHz
4.75 GHz
3 GHz
Global primary MS band
Non-global primary MS band
Non AI1.13band
3.25 GHz
24.25
27.5
29.5
31.8
33.4
37
40.5
42.5
43.5
5.25 GHz
0.85GHz 7GHz
ITU
• Global harmonized spectrum for high band can boost 5G infrastructure,
terminal and semiconductor economy scale and speed up emerging service.
China
USA
Japan
Korea
EU
Taiwan
1.6GHz 5.5GHz
6.5 GHz 7GHz
unlicensed
Considered for 5G by global regulators
Further Studied by MOTC, Taiwan
40
30
45 50 60 70GHz
0.6GHz Sharing Spectrum
MOEA 5G R&D Spectrum(38.6-40GHz)MOEA 5G R&D Spectrum (28GHz)
Source : 5GIA, FCC, IMT-2020(PG), 5GMF, 5G Forum, and MOTC Taiwan, Nov. 2017. Edit by
5GO and MIC.
8. 5G Commercial Deployments
•Fixed wireless, router, smart phones
•Islands of deployment without continuous national
coverage
•Non-standalone (NSA), EN-DC architecture available,
SA is yet to be proven
•eMBB as initial services
7
13. Demands for Enterprise Network
• Secure
• Avoid operating confidential data outflows or external
security attacks (private network vs. public network)
• Ownership, QoS control
• Total controls of network access, traffic quality (security、
QoS)
• Reliability and flexibility
• dedicated network、not subject to other interference
• Coverage
• Applications tailored end-end solution
• Quickly import professional domain knowledge and fast
deploy innovative service
12
14. Taiwan 5G Activities
• 5G Program Under MoEA
• ITRI/III, domestic and international partners
• Target for vertical applications, trials for factory operation, smart hospital,
chemical plant and entertainment
• EU-Taiwan collaboration
• H2020 EU-TW Open Target Call Phase II: Clear 5G, 5G Coral
• Phase III projects: 5G CONNI, 5G DIVE to be started in Oct. 2019
• Taiwan 5G alliance
• Lead by operators
• To facilitate collaboration among MNOs and domestic ICT industries on
small cells, switches, servers, terminals, and IoT devices
• Pilot team led by CHT was formed in Jan. 2018
• Pioneering team led by FET was formed in Nov. 2018
• MoST/MoEA joint R&D
• To leverage resource and research results from academia and apply results
to system implementation
13
15. 5G E2E System Architecture
700MHz~70GHz
Licensed +
Unlicensed
5G NR
gNB
HSS PCRF
4G
S/PGW
Core Network
MME
Transmission
network
2x2 antenna
20MHz BW
LTE
150Mbps
BBU
RRH
eMBB UDN URLLC IoT
5G
Transmission network
Transmission
network
BPN
SON Server
UDN Server MTC Server
General Purpose
CPUvEPC vEPC
vEPC/5GC
BPNCU CU
SON Server
iMEC GW
Light vEPC/vNGC
UHDVR Sensors
Transmission
network
CU
mmWave
5G NR
gNB
Transmission
network
drone vehicle
Network Slicing
NFVI
NFVI
General Purpose
CPU
Network
virtualization
14
5GC
17. Taiwan 5G Spectrum Plan
16
~1GHz 1~6GHz 6GHz~
Phase 1
(Early 2020)
1775~1785/1870~1880 MHz
3.3-3.57 GHz
27~29.5 GHz
Phase 2
(Planning)
600 MHz
800 MHz
4.5~5.0 GHz
24.25~26 GHz
37~40 GHz
Spectrum
Policy
Trial
MVNO、MEC and Network Slicing
Vertical Industry can lease spectrum from
telecom operators
Telecom
Operator
Evaluating dedicated spectrums for private
networks for security and performance
purposes.
Private
Network
Provide innovative experimental spectrums for
5G PoC or PoB trials before 5G spectrum auction
Enabling 5G Vertical Application
Source: Taiwan 5G SRB, Oct. 2018
18. Summary
•Global spectrum harmonization is in the right
direction
•Initial commercial launches are mainly for eMBB &
fixed wireless, isolated/dense hot spots
•Fully service based infrastructure is yet to be
architected and implemented
•For vertical applications
• Business model, architecture, spectrum, flexibility,
specifications, performance etc., need to be addressed
17
20. 5G Pre-commercial Trials
8
Immersive viewing experience at first row seat with
VR 360 headset
The 1st real-time VR 360 live streaming show adapting 5G eMBB
Enterprise Network of Taiwan Industries
Arbitrary view points
switching, favorite
perspective
to catch exciting moment
Share fun with friends
through VR interactions and
enjoy real-time body
expressions
21. The Next 5G Field Trial
- Chemical Plant
Display
route for
inspection
Real time
display of the
inspection
results
Automatic recording of
the inspection results
AR goggle display
SoP w/remote
assistance
5G+AI Cam for Face
Recognition on Smart
Glasses
Access control Operation Safty Assurance
Site Inspection w. AR Goggle Streamline Operation Efficiency
5G
120
22. Smart Hospital
Health Monitoring &
Tracking
Position
tracking
Medical devices,
equipment tracking
Remote Diaganosis
5G connection
High resolution video for remote
assistance
Optimize medical resources among hospital
system
Via AR for pre-
surgerical/medical
training, or remote
assistance
4k/8k Low Latency AR/VR
Data collection
25. Deployment Scenarios
24
Public
Wide
area
Low Entry
point
Spectrum RAN
Core
Netwrok
Enterprise
Control
Center
MVNO
Mobile Virtual
Network Operator
MVPN
Mobile Virtual
Private Network
Private
Network
control
Security
Regional
Private
Network
: Enterprise: Telecom Operator
26. European Conference on Networks and
Communications (EuCNC) 2019
European and Taiwanese Cooperation on 5G
Valencia, 19 June 2019
Bernard Barani
Deputy Head of Unit, Future Connectivity Systems
European Commission – DG CONNECT
27. Motivation for International Cooperation in Network
R&I
• Industry Drive, competitiveness across value chains;
• Pre competitive opportunities, towards
• Global consensus and standards;
• Interoperability, spectrum
• Use case best practices
• Addressing global societal challenges
• Specific policy issues, fn(partner country), e.g. reciprocity
28. 5G Public Private Partnership
Industrial R&I cooperation cartography
29. R&I , 2018-20 Work Programme Number of
Projects
Total N° of
projects *
Partner
- Applications and trials with 5G networks
- Beyond 5G, applicability of spectrum >275 GHz
➔spectrum, interop, use cases, Beyond 5G
2 6 5GMF
- Application trials at mmwave bands
- Interworking across multiple radios
➔ Standards validations, use cases
2 4 5G Forum
- eMBB Trials at 3,5 GHz and in the V2X context
➔ standards, 5G V2X
1 1
IMT 2020 (5G)
Promotion
Group
- 5G trials addressing end-to-end testbeds for specific
applications
➔ 5G verticals,
2 4 DoIT-MoEA
- Coordination of EU-NSF projects relevant to the
Advanced Wireless Platform programme
➔ Longer term beyond 5G
1 1 NSF
*5G/Network related only, under H2020 programme
30. EU-Taiwan: Horizon 2020 and 5G as catalyser
5G PPP Phase 1: Classical cooperation
5G PPP extension: Two dedicated call for Taiwanese partners
Two EU-TW call 1 projects
started in September 2017
34. Opportunities for the next decade
« Digital Industries » « Physical Industries »
Share of GDP 30% 70%
Digital Investment 70% 30%
Annual Productivity
Growth (15 Years avg)
3% 0,7 %
Automation and Industry: 3,5 to 10 Trillion € by 2025, 11% of
economy (Mc Kinsey)
Network share prospects, 10%? Doubling current broadband revenues?
➔ Assumption 1: Industrial/Vertical applications will remain a strong
innovation driver over next decade
Source, Nokia quote from:The coming productivity boom, Michael Mandel, Brett Swanson
35. What we observe today:
- Social issues, coverage
- 3,5 Billion people without wireless Internet
- Energy, sustainability in hyperconnected society
- « Energy skyrocketing at the edge ».
- EMF raising concerns
- What impact of untested spectrum usages? How to decrease exposure?
- Human centricity and trust, data control and governance
- Security and Autonomy
- Coping with embedded critical infrastructres
➔ Assumption 2: Societal issues to gain accrued importance
36. 5G Vision and focus Parameters: will they remain valid?
11
Use cases and drivers
• Capacity, still 50% traffic increase/ year
• local applications, sub-ms latency
• Gbps availability, e.g XR applications
• Extreme reliability beyond 5x9;
• mMTC “everywhere “
• Extreme energy efficiency
• Further enhanced high security/trust
• Very high mobility
• cm-level localization
Source: ITU-R Rec.
M.2083 (modified)
Enhanced mobile
broadband (eMBB)
Massive machine type
communications (mMTC)
Ultra-reliable and
low latency
communications
(URLLC)
Network traffic
(exabytes/mont
h)
Positioning
accuracy
Security
Distributed
computing
Smart Networks &
Services
37. Assumption 3: 5G design parameters pushed towards new frontiers
will remain valid towards wide industrial applicability
Disruptions may be expected, for example:
- Innovative spectrum use towards sensing and environment
augmentation
- Generalised use of AI and Machine Learning in multiple aspects,
Intelligence and semantic
- Multiple network architectural issues (extreme agility, energy, blurring
device/network/cloud, security)
- Untested technologies at scale, e.g. blockchain
38. Horizontal issue: Energy Efficiency
- Energy needs, significant increase since 2014,
expected to accelerate;
- by 2030, 10 M edge clouds , 9 M robo-cars/yr ➔
new architectures.
- Optical, virtualisation, densification: parts of the
solution
- Other techniques, energy harvesting and ambient
energy use
➔ Towards EE as part of the network
management, « EFCAPS » + E2E integration
Source: Anders Andrae « best case », Nature News Feed,
Sept 2018.
39. Unified &
Access-
agnostic
Authenticat
ion
Primary
Authenticat
ion
Secondary
Authenticat
ion
Increased
Home
Control
Initial NAS
Security
&
Privacy
Visibility
and
Configurabi
lity
Service
Based
Architectur
e
Steering of
Roaming
5GS – EPS
Interworkin
g Security
LTE-NR
Dual
Connect.
(Option-3)
PLMN
Interconne
ct Security -
SEPP
RAN
Security –
DU-CU Split
Network Slice Security
Long Term Key Update
256-bit Algorithms for 5G
KDF Negotiation
Vertical services and LAN
Single Radio Voice Continuity
from 5G to UTRAN
Wireless and Wireline
Convergence Security
Cellular IoT Security for 5G
5G Phase I 5G Phase II
“Journal of ICT Standardization” OpenAccess by River Publishers Special issues on “5G non-standard aspects” and “3GPP 5G specifications”
Beyond 5G?
> SaaS
Interoperability, E2E
Quantum
AI based malware detection
GDPR
(Multiple) Identities
Horizontal issue: Security
Cross domains blockchains
………………
40. Assumption 4: Significant advances compared to foreseeable 5G
will come from the combinatorial effect of a multiplicity of
technologies, use cases, societal requirements, and business
models.
➔ A modified approach may be required from the start.
42. Devices: Multiplicity of
Connected Devices Industrial
Automation 360o VR/XR
Fully Automated
Vehicles
Haptic
Communications for surgery
Agri sensors
Smartphones Computers
Mobile & Last
Mile
Networks
Versatile Infrastructure
Multiple Topologies High
density
access
Corporate
nets
Indoor Short
Range
Dense
IoT
Fixed wired
access
OLT
CU
AWG
RU
RU RUDU
Service provisioning
Computing and Storage
Data Analytics
End-to-endResource
ManagementandEnergy
Efficiency
Drones
Requirements
Industrial & Consumer
Applications Telemedicine Construction Connected
Mobility
Environment Factory Immersive
tourism
8K movie Sport &
events
…..
End-to-endSecurityand
Trust
Smart Networks and Services - Value Chain Approach
New opportunities
Enabling Technology
Components
43. Beyond 5G: A Possible Roadmap
2022 20242020 2026 2028 2030
Un-constrained
R&D
6G Design
R&D
Standards
SI launch:2006
Trials Launches
Derived from Orange
~ Based on modified 5G Model
Need agility in case of accelerated commercial pressure
R&I International
cooperation window?
44. Timeline and process for the preparation of
Article 185/187 initiatives
3 May – 27 June: Structured consultation of Member States (as part of strategic
coordinating process)
May-June: Publication of draft Inception Impact Assessments and start of the Impact
Assessment work
Mid-June until Open Public Consultation on future European Partnerships based on
Article September: 185/187
July SNS Stakeholders Workshop (extended, tentative)
24-26 September: European R&I Days (policy discussion and validation with stakeholders,
covers all European Partnerships)
October SNS Stakeholders Workshop
End of 2019: Submission of Impact Assessment drafts to Regulatory Scrutiny Board
Early 2020: Adoption of Commission proposals for Article 185/187 initiatives
Early 2020 Finalisation of SNS SRIA and Roadmap (TBC)
Early 2021: Launch of first European Partnerships under Horizon Europe
Indicative timeline European Partnerships
45. By Way of Conclusion
The B5G-6G journey has started. At this early stage, flexibility is key. We can today identify
as potential drivers:
• Networks in industrial environments, pushing the 5G envelope limits
• Societal issues to get enhanced focus
• 5G design drivers duly complemented remain valid
• New innovation/disruptions to be integrated (AI/ML, DLT, mEC..)
• New system/stakeholders approach targeted
• Europe committed to support EU excellence in this critical domain.
47. Clear5G
Communication for the Factories of the Future
Klaus Moessner (Technical University Chemnitz, University of Surrey)
Project Coordinator, H2020 Clear5G project
EuCNC 2019
WeC8: Workshop 6 : European and Taiwanese Cooperation on 5G
Valencia, 19 June 2019
48. Project Vision:
“to provide technical solutions that enable
future 5G networks to act as dependable
communications backbone for the Factories
of the Future”
Valencia, 19.06.2019 2
49. Project Brief
• Clear5G-Converged wireless access for reliable 5G MTC for factories of the future (FoF)
Valencia, 19.06.2019 3
• Horizon2020 EU-TW collaboration
• Sept. 2017 – Feb. 2020
• Focus on 5G radio network for FoF (PHY, MAC, NET)
• Coordinator: University of Surrey
• Coordinator TW: Institute for Information Industry
• Technical Manager: TNO
• Objectives: to design, develop, validate, and
demonstrate an integrated convergent wireless
network for Machine Type and Mission Critical
Communication (MTC/MCC) services for Factories
of the Future (FoF)
50. Clear5G network view
• While parts of the underlying communications infrastructure will be public networks, within factories there
will be private (physically or virtually) factory wide 5G networks tailored to the particular needs of the
individual site
• Continuous monitoring while products or parts are within the logistics section of the production chain
• Spectrum regulation and management plays a significant role
Valencia, 19.06.2019 4
51. Challenges and Clear5G KPI’s
The Industrial environment is challenging for
wireless connectivity, e.g. both large- and small-
scale fading (predictable?)
• Massive connectivity
• Coverage, reliability and latency
• Heterogeneity (private and public network,
different radio technologies)
Valencia, 19.06.2019 5
Clear 5G KPI Targeted value
Latency (end-to-end) Down to 1 ms
Reliability Up to 99.999%
Connection density Up to 100 nodes per 1 m3
Security PHY framework
Heterogeneity (convergent
air interfaces)
Coexistence of various radio
interfaces, and various FoF
use cases
Energy efficiency (Device
battery life)
>15 year battery life
52. How are we getting there?
The technology components
• PHY (WP2)
• Adaptive frame structure
• New waveform
• Non-coherent modulation
• NOMA
• Physical-layer security
• MAC (WP3)
• Random access enhancement
• (Adaptive) Contention-based or –free MAC
• Joint PHY and MAY optimization
• Heterogeneous Radio Access
• Networking (WP4)
• RAN architecture
• RAN Slicing
• Multiple connectivity, (UE) relaying
• Public and private network integration
Valencia, 19.06.2019 6
53. Example: Hardware prototyping of sparse code multiple
access (SCMA) for massive connectivity in FoF
SCMA: A code-domain NOMA which
can support massive connectivity by
efficiently exploiting the sparsity of
codebook using message passing
algorithm.
Objective: To implement and
demonstrate SCMA system over
USRP testbeds.
Testbed hardware & software
• One NI-PXIe
• Two USRP RIO-2943R
• One CDA-2900 (10 MHz
frequency clock)
• LabVIEW Communication
System Design
Valencia, 19.06.2019 7
RIO1 TX
PXIe
RIO2 RX
Channel
Host PC
System parameters Values
Center frequency 2 GHz
Bandwidth 10MHz
FFT length 64
CP length 8
# of RB per frame 200
# of samples per frame 12000
54. Example: Network slicing in a factory network
• Slicing enables operators to support different
network instances on the same infrastructure
• FoF as one of the slices, or
• Different FoF use cases (e.g. URLLC, non-
URLLC) may be served by different slices
• URLLC: local controller
• Non-URLLC: controller in the cloud
• FoF slices may be provided by a public
network operator, or a physically private
network.
Valencia, 19.06.2019 8
55. … leading to Traffic Abstraction and Analytics
Showcase managing wired FoF network wrt
specific requirements and analysing traffic
statistics gathered from the network
Steps to follow:
✓ Define multiple traffic classes in
wired network
✓ Create and install a set of wired
paths for each traffic-class – road
network
✓ Mix traffic classes so as to achieve a
fair distribution between different
traffic classes
✓ Collect traffic statistics and analyse
them
Valencia, 19.06.2019 9
SDN
controller
Systems room
Factory area: 50x50m
NG core
URLLC ctrl
1
URLLC ctrl
2
non-URLLC
controller
URLLC area 1
URLLC area 2
system mgr
non-URLLC area
RAN ctrl
orchestrator
edge core
UPF UPF
56. Example: Closed-loop control of industrial AGV with UE
relaying support
Objectives
• Demonstrates that 5G technologies can fulfil the strict
requirements of a close-loop controlled Automated Guided
Vehicle (AGV) moving in a factory.
• Support low latency exchange of data in monitoring and
analysis services such that the AGV can be remotely guided
based on product quality results.
• Showcases UE relaying in a factory environment as a mean to
improve reliability.
• If the default communication link of the AGV is unavailable,
the AGV will use nearby UE(s) as relay node(s) in order to
reach the destination node.
Technical Benefits
• Low latency exchange of data between the
industrial devices (e.g. AGV, factory server)
• Improvement in reliability by using multiple
connectivity options
• Improvement in radio coverage by using UE
relaying
• Improve the level of industrial automation
Valencia, 19.06.2019 10
Processing Quality Check Packaging
AGV (equipped withUSRP
for5G connectivity, Wifi/LTE
multi-connectivity)
QC Passed
ImproperMaterial
Cloud platform (located in the Cloud
or in a central factory location)
USRP PC
•Analysisof quality report
•Decisionon AGV route
•AGV control
•SLAM (SimultaneousLocalizationand Mapping)
•Navigation
•Quality checking (e.g.using the camera,the
dimensioning sensors)
•Sendsquality report to the Cloud platform
•Receivesrouting from the Cloud platform
•Acts as a relay node
•Forwardstraffic to server
USRP (OAI)
AGV
UE (OAI)
USRP (OAI)
BS (eNodeB, gNB)
EPC (OAI)
57. For more details, visit the Clear5G
booth in the exhibition area!
Valencia, 19.06.2019
11
58. Beyond simulations: the trial scenarios
• Massive sensor data collection in FoF
• Monitoring & Closed Loop Control in FoF
• Seamless Interoperability and Mobility
(private FoF and public networks)
Valencia, 19.06.2019 12
60. 1. Machine 3. Cleaning
2. Polishing/
Buffing
A+ Project Data
Collection
Data from
the motion
of the arm
WINGS
AGV
Clear5G
Robot
Control and
decision
1. AGV Control (WINGS) ok
2. UE relaying/multi-
connectivity (TNO) ok
3. Massive sensor data
collection (III) ok (A+)
4. Packet duplication (UNIS)
→video
5. LORA/LPWA (CEA) ok
6. Slicing (ARG/TT) ok
~50/60m
What is 5G?
• UE Relaying and RAT Selection
• Communications with AGV (~40ms)
• Slicing
Cloud
1
2 3
4
App
Server
LTE & UE
Relaying
5
7
Human intervention
(e.g. loading/
unloading from AGV)
Clear5G
factory setup
6
Quality
checking
(Discard or
continue)
Pillar
Pillar
Intranet
61. 15
This project has received funding from the European Union’s Horizon 2020 research and innovation
programme under grant agreement No 761745 and from the Government of Taiwan.
Thank you!
More info: http://www.clear5g.eu or follow Clear5G on twitter. @Clear5G
64. Introduction - The Cloud, Edge and Fog
2
User
Telco
Network
Cloud
APP
Cloud
User
Telco Network
Cloud
Edge
APP
Edge FogProximity Distribution
APP
User
Telco Network
Cloud
Edge
APP
APP
Fog
APP
Fog
APP
Fog
APP
User-
Centric
65. 5G-CORAL Project Vision
• 5G brings a flexible RAN architecture including flexible functional split of the
gNB/eNB (DU <-> CU) and a diverse set of terminal (UE/CPE) types which all
have computing capability ready for harvesting
• Edge and Fog are complementary, and jointly together will define the computing
substrate of next generation radio access networks
4
The computing fabric does not stop at the Edge. It is more pervasively distributed through
the Fog into access infrastructure and user terminals
UE
CU
CU
Core
Data
Network
CloudCloud
DU
DU
UE
UE
Edge
DC
…
FogFog
A rich pool of
distributed
computing
fabric
Fog
66. 5G-CORAL Project Mission
5
Target an integrated virtualized solution deep into the RAN offering distributed data
services for various applications
Mission: To develop
the framework for an
integrated virtualized
Edge/Fog solution and
demonstrate its value
proposition for various
use cases
67. Solution Building Blocks
• OCS: A logical system
for composing,
controlling, managing,
orchestrating, and
federating one or more
EFS(s). An OCS may
interact with other OCS
domains
• EFS: A logical system
providing service
platforms, functions and
applications on top of
Edge and Fog resources.
It may interact with other
EFS domains
7
Builds on two sub-systems, the EFS (Edge and Fog computing System) and the OCS
(Orchestration and Control System)
68. Integration and Demonstration
8
EFS and OCS baseline implementations integrated and demonstrated at several events
Use Case Demonstrations
Augmented Reality Shopping Mall (Nov’18)
Virtual Reality Taiwan’s UK visit (Sep’18)
Edge Congress (Sep’18)
EuMA (Sep’18)
Shopping Mall (Nov’18)
Fog-Assisted Robotics EuCNC (Jun’18)
5TONIC (Jun’18)
EuMA (Sep’18)
Shopping Mall (Nov’18)
IoT M-RAT Gateway EuCNC (Jun’18)
Shopping Mall (Nov’18)
Connected Cars Torino Trial (Oct’18)
69. 5G-CORAL Evolution: 5G-DIVE
■ Goal: Design, validate and verify an intelligent 5G solution that
integrates 5G connectivity with edge and fog computing (and
intelligence residing on this new distributed edge).
■ Target products: UE (including, drones), CPE, gNB, fog, edge, and
core
9
Europe Taiwan
70. Vertical Pilots – Industry 4.0
Digital Twin Apps
• Provides a virtual replica of a robot or of a part of
a production line.
• The 5G network coverage will be deployed to
enable real-time visibility and remote insights into
robot status and performance without having to
directly operate on the physical machine.
• Requires eMBB and URLLC for the on-time delivery
of the information of the sensors to the virtual twin
and for the interaction with the digital model.
• Facilitates assessing the concepts of remote control,
monitoring for preventive maintenance, and safety.
Connected Worker Augmented Zero Defect
Manufacturing (ZDM) Decision Support
System (DSS)
• Explores the capabilities of Fog/MEC/Cloud multi-tier
Edge to address this local processing and visualization
of geometric features for manufactured parts.
• Deploys in the Fog devices (e.g., video cameras),
algorithms able to detect characteristic patterns for
defects in the production.
• Requires eMBB for the interaction with the platform for
reinforced learning and URLLC for processing of results
in the Fog devices.
• ZDM techniques may potentially reduce scrap by
100%, and predict form and/or welding errors.
10
71. Vertical Pilots – Autonomous Drone Scout
Drone Fleet Navigation
• Improves current Drone product portfolio,
enabling a better piloting of the Drone swarm.
• Providing intelligence in the Drones
• Requires eMBB and URLLC for the on-time
delivery of the information of the sensors to the
edge data centre for drone interaction.
• Enables new Drone-based services:
• delivery, inspection and monitoring,
scouting, Aerial Imaging, and precision
agriculture on large scale.
Intelligent processing of images in the
Drones
• Enables the deployment of intelligent
functions in the Drones and its cooperation
with the different tiers of the 5G-DIVE
platform.
• Requires eMBB and URLLC for the on-time
delivery of the information of the Drone.
• More automation in the scouting processes,
creating a new value chain of services
which can be used to provide more services
to the customers.
11
72. What are we missing in 5G-CORAL?
LatencyHigh Low
Data FilteringLow High
LocalizationCentralized Distributed
73. 5G-DIVE Building Blocks
VIM
EFS Manager
EFS Orchestrator
EFS Functions
EFS Applications
EFS Service Platform
Operation Support System Business Support System
5G-DIVE Elastic Edge Platform (DEEP)
EFS OCS
Infrastructure and platform systems
Vertical industries systems
Business Automation
Support Stratum
hosting all proposed virtualized functions, services, and applications
managing and controlling the EFS, and its interworking with other domains
(1) EFS:
(2) OCS:
supporting vertical industries in day-by-day operations, management, and automation
of businesses processes on-top of an edge and fog infrastructure.
(3) DEEP:
13
74. Consortium partners and acknowledgment
06 December 201815
www.linkedin.com/in/5g-coral5g-coral.eu twitter.com/5G_CORAL
Cost:
3.856.973,75€
Technical Managers:
Dr. Alain Mourad (Interdigital Europe), Dr. Tony Do (ITRI)
Coordinator:
Dr. Antonio de la Oliva (UC3M)
Project lifetime:
01/09/2017 - 31/08/2019
The 5G-CORAL projet has received funding from the European Commission H2020 Grant No. 761585
88. TONIC Research Group
Applying NOMA for Latency Reduction
in Factories of the Future
June 19, 2019
Hung-Yun Hsieh
Graduate Institute of Communication Engineering &
Department of Electrical Engineering
National Taiwan University
90. TONIC Research Group
Clear5G – WP3
3
• WP3 provides MAC-layer air interface enhancements
for the MTC in the FoF use cases, including both the
control plane and the user plane.
• WP3 investigates cross-layer (PHY and MAC) IoT traffic
management and the impact of heterogeneous radio
networks.
• Task 3.1: Random Access Enhancement
• Task 3.2: Adaptive MAC Protocol for mMTC and uMTC
• Task 3.3: Heterogeneous Radio Access
• Task 3.4: Implementation
92. TONIC Research Group
Clear5G – Task 3.2
5
• Task 3.2: Adaptive MAC protocol for mMTC and uMTC
• Task 3.2 proposes an adaptive MAC protocol, which
works in a contention-like manner in low-load traffic
conditions to reduce the access latency. Moreover, the
proposed MAC protocol will work in a scheduling-like
manner in high-load traffic conditions to increase the
system throughput and reliability.
93. TONIC Research Group
Clear5G – Roles of III and NTU
6
• III (task leader) contributed to the design of MAC
scheduler for massive connection, in addition to
exploring designing flexible MAC concepts for
reconfiguration of the amount of resources for RACH,
system signaling, scheduling of low-latency users, etc.
• NTU proposed designs of low latency random access
for massive MTC, with the coexistence of multiple
radio access technologies.
94. TONIC Research Group
III Contributions (D3.2)
7
• III has designed a MAC-layer mechanism that
performs resource allocation to users adaptively
either in OMA or NOMA.
• Code-domain NOMA (SCMA or LDS) and OMA (OFDMA or
SC-FDMA) uplink multiple access schemes.
• F candidate UEs to be allocated either in F resource blocks
(OMA) or V>F resource blocks (NOMA)
Omax: max modulation order
NOMA BLER based on the worst SNR
• Choose the MA with the best throughput (SOMA or SNOMA)
95. TONIC Research Group
III Contributions (D3.2)
8
• Evaluation results with low and high SNRs
• Low SNR: UE power = -20dBm
• High SNR: UE power = -10dBm
• Latency of the proposed mechanism can always
achieve the lowest irrespective of the conditions
Low SNR High SNR
96. TONIC Research Group
NTU Contributions (D3.2)
9
• Two-tier architecture with heterogeneous radios
• Tier-1 (to/from base station): 5G
• Tier-2 (between IIoT devices): short-range radio (6TiSCH)
• Latency reduction
• Pure 5G -> two-tier: reduce bottleneck at the base station
• Pure 6TiSCH -> reduce multi-hop delay
• Use NOMA at tier-1 can further
• reduce latency
• Some interesting results for
• IIoT devices
97. TONIC Research Group
Data Characteristics of IIoT Devices
10
• IIoT devices are deployed to collectively gather (or
report) data required by the target application
• Collected data by individual devices is often related
• Conventionally traffic from individual UEs are considered
independent
• Contention of radio resource
• Conventionally, it matters to provide QoS to each UE
• For IIoT devices, it is possible that they contend resource to
transmit same (similar) data
• Does it make sense to provide QoS to each IIoT devices
without considering the data they carry?
• Consider the problem of pairing & scheduling NOMA
UEs
98. TONIC Research Group
Sum-Rate Maximization Scheduling
11
• Conventional approach for NOMA user pairing is to
maximize the sum data rate of the pair
• machines are indexed in decreasing channel gain
• Achievable rate of each machine
• P is the transmission power of each machine
• G accounts for the spectral gap to Shannon capacity
100. TONIC Research Group
Resource Minimization Scheduling
13
• Sum rate maximization is typically good for
backlogged traffic sources (e.g. FTP), yet is it the case
to M2M communications used for data gathering?
• Resource (time) needed for machine k
• is the size of data
• Required resource for a given pair
• “Faster” machine needs to wait for the slower
101. TONIC Research Group
Waiting-Time Minimization
14
• Waiting time inside a time slot indicates a waste of
radio resource
• Maximizing resource utilization is beneficial to
resource minimization
• Scheduling metric
• Scheduling pairs in
decreasing metric value
102. TONIC Research Group
Overall Comparison
15
• Radio resource minimization and waiting-time
minimization achieve the desired performance
103. TONIC Research Group
Summary
• Reducing latency in FoF using NOMA
• Adaptive MAC to choose the best multiple access
scheme depending on the achievable throughput
• Two-tier architecture provides more flexibility with
further reduction in end-to-end latency
• For IIoT devices with limited amount of data to send,
throughput maximization does not necessarily lead to
latency minimization
• Better user NOMA pairing and scheduling methods
can be designed optimize latency performance in FoF
16
104. 5G CORAL Distributed Edge and Fog Computing
Network Infrastructure and AR Navigations
Dr. Jen-Shun Yang
Manager of Advanced Communication Technology & Standard Development Dept.
Division for Video & Multimedia Communications Technology
Information and Communications Research Laboratory
Industrial Technology Research Institute
105. Content
• Motivation of Distributed Computing Resource Orchestration
• 5G CORAL Use Cases
• Introduction to Fully Distributed Computing Resource
Orchestration
• AI CNN Training for Resource Allocation Optimization
• Performance Testing
• Conclusions
2
106. 5G CORAL
Hierarchical Multi-tier Computing Infrastructure
3
5G
Edge&Fog
Control
Plane
5G
Edge&Fog
User Plane
5G
Edge&Fog
Hierarchical
Multi-tier
Computing
Infrastructure
and Use Cases
A 5G Convergent Virtualized Radio Access Network Living at the Edge
107. • Complex 3D indoor
environments in train station
• Imagine that you are lost in a Taipei
Main Station, looking for the entrance
to MRT/HSR/TRA/Bus station/taxi
ranks and up to 70 exits to parking and
main roads.
• Or you are trying to find a specific
department or restaurants in an
enormous shopping malls.
• Since GPS is not working indoors, do
we have any other better choose, except
for the existing i-beacon and Wi-Fi
positioning systems?
4
3D Taipei Main Station MAP
Motivation: Indoor Navigation in Taipei Main Station
108. ❖Taiwanese manufacture ASKEY Computer was requesting ITRI to provide
5G CORAL Fully Distributed EFS&OCS and AR solutions for AR
Navigation and Advertisement services in Taipei Main Station.
❖Video of an exemplary AR Service Scenario in Train Station Shopping Mall
5G CORAL Use Case: Distributed Computing for
Indoor AR Navigation/Advertisement
5
⚫ Low End-to-End (E2E) Service
Latency to fulfill better user
experience (e.g., < 1sec)
⚫ Support High User connection
Density
⚫ Heavy Computing Loading
shall not be happened in
user’s smartphone
– Small Size (database)
User APP
– Low power consumption
⚫ Indoor Localization precision <
1 meter
Performance Requirements
109. Distributed Computing for Heavy Load AR Recognition
6
Fully Distributed Computing Resource Orchestration Benefits
Low Latency (E2E): Minimize computing latency by processing image recognition and navigation tasks at the EFS
Connection density: Increase number of connection by distributing the incoming requests
Service Reliability: Overcome Single Point of Failure (SPF) by centralized distributed computing mechanism
Crashed!!
Centralized Load
Balanced Controller
Non-Centralized Load
Balanced system
SPF
Crashed!!
FCD
FCDFCD
FCDFCD
FCD
FCD
FCD
Traffic
Jitter
FCD: Fog Computing Device
110. Optimization of Fully Distributed Computing Resource Orchestration
• AI CNN offline trains the dispatching model of computing requests for Distributed FCD Cluster.
• Obtain online load balancing optimization among Distributed FCDs in Cluster.
FCD
FCD
FCD
FCD FCD
FCD
CNN offline Training
with
Fully Distributed
Computing
Resource
Orchestration
AI SoC Module
1 2
43
65
7 8
7
Distributed FCD Cluster (No:1~6)
No Central Controller
• Network Capacity Constrain
• Delay Constrain
• Network Topology
• Computing Capacity
• Load Balanced Constrain
Training Input Data
Trained Model in each FCD for
Resource Allocation Optimizer
(request dispatcher)
Decision time range:
0.3ms~1ms
HPF: High Performance
FCD: Fog Computing Device
HPF
PC
FCD
3 1
25
6
4
111. 8
Performance Testing Configuration for AR Image Recognition in
Distributed Fog Computing Cluster
preview
camera
send
images &
beacon ID
features
extraction
features
detection
features
query
Target
determine
object
rendering
UI Display
local
images
Web CAM/Smart Phone/AR Glasses
E2E Low latency < 500ms
AR Image Recognition Service Flow
Projector/Smart Phone/AR Glasses
nVidia TX2 FCD
112. Distributed Fog Computing with Native
AR application
• Using TX2 as Fog CDs without Virtual
Machine
• The E2E latency, measured by Timers
1, 2,3 and 4, is 279ms.
• Including the time for the Distribution
Handler to update the Statistics with all
of the databases(step 2, 3), the time
cost by the Resource Allocation
Optimizer(step 4, 5) and the time for the
Distribution Handler to read the
Distribution Decision result and dispatch
the job to AR Server(step 6, 8).
9
Virtualized
Software Module
Distributed Fog CD Manager
Design
113. 10
• Using TX2 as Fog CDs with Container
– User Plane virtualized by LXD
– Control Plane signaling by FogO5
• The E2E, measured by Timers 1, 2, 3
and 4, is 317ms.
Distributed Fog Computing with Container based
AR application
Virtualized
Software Module
Distributed Fog CD Manager
Design
114. Fog Computing Device Software Platform Spec.
Have developed modules and prototypes:
• Fog Virtualization Layer Solutions
• Docker/LXD virtualized container technologies
• Fog Middleware Solutions
− RESTFUL API
− D2D based Wireless Relay Mesh Networking with
Fast Deployment SON, Smart routing, and
Broadband Relaying
− Fog Plug&Play Configure Agent
− Fog Application Monitor Management Agent
− Fog Parallel Computing Control
− Resource Discovery Agent
• Fog Application and Service Prototypes
− AR Navigation
− Smart Lamp Pole (S-L-P) Video Surveillances and
Recognition
can identify/recognize illegal parking on red lines, roadside
parking spaces
− IIoT Robot Fog Computing
− Stadium VR360 Broadcast Multicasting
OK
Under
Developing
Open
Source
Module Status:
IIoT Robot Fog Computing
SoC
Drone Fleet Flight Control
for Fast Network Deployment
Augmented Reality Indoor &
Outdoor Navigation
Stadium VR360
Broadcast/Multicast
Internet of Things Cloudlet
Hotspot
S-L-P Video Surveillances
and Recognition
Fog
Application
and Services
Fog
Virtualization
Layer
Hardware Platform
Wi-Fi/LTE cV2X
Container
RAT
D2D Wireless
Relay Mesh
Networking
Fog P&P
Configure Agent
Fog
Distributed
Computing
Service
Migration
Resource
Discovery
Agent
Fog Application
Monitor Agent
Fog Security
Management
VIM
interfacing
Container Container Container Container
Application Support
API API API API API
Fog
Middleware
Application Services
NA
OCS clients
11
115. Fog Computing Device Hardware Spec.
12
Fog Computing Device (FCD)
Reference
Price $720 Euro
116. Conclusions
• Technology Objectives of Fully Distributed Fog Computing System
• Industry Impacts of Fully Distributed Fog Computing System
– New Solution for Indoor Navigations
– New Business model by AR Advertisements
– Successfully done the technical transformation to Taiwanese Manufactory and
ongoing deploying in Taipei Train Stations
– Potential Technologies for the Fog Computing of Car Fleet and UAV Fleet
24 June 201913
118. preview
camera
send
images &
beacon ID
features
extraction
features
detection
features
query
Target
determine
object
rendering
UI Display
local
images
Web CAM/Smart Phone/AR Glasses Wi-Fi EFS in FCD
E2E Low latency < 500ms
Wi-Fi AP #n
FCD#n
Etherswitch
RJ-45
RJ-45
Wi-Fi AP #1
FCD#1
Service Flow:AR Video Recognition
Projector/Smart Phone/AR Glasses
Ultra low latency Load
Balanced Distributed
Computing Cluster
Enterprise Cloud
Internet Cloud
OCS
Monitoring
Distributed Fog CD Cluster
Deployments of Distributed Computing EFS and OCS
Filtered by
i-Beacon
i-beacon
Internet Cloud AR Computing
E2E Low latency 5 sec
Enterprise Cloud AR Computing
E2E Low latency 1.5 sec
EFS: Edge and Fog computing system
FCD: Fog Computing Device
OCS: Orchestration Control System
15