AIDRC Gitex Africa - Final edited_TL.pdf

tii.ae
Paving The Way to NextGen Sustainable
Digital Cities Infrastructure
Disruption With AI and B5G/6G Networks
Thierry Lestable, Ph.D, Dipl-lng.
Executive Director,
Artificial Intelligence and Digital Science Research Center

2
TII – Technology Innovation Institute
Combating Climate Change
The Need for Smart and Sustainable Infrastructures!
Current level is at 1.2 C. finding the global solutions to reduce
emissions by 45% by 2030 and reach net
zero by 2050.
In Europe, transportation is the main source of carbon emissions:
- It represents 27% of total emissions in the EU
- 45% of these emissions come from cars.
Road transport is the main contributor to these emissions: 77%.
In France: 121 million metric tons of CO2 in 2018.
The European Green Deal aims at reducing the emissions of
the transport sector by 90% by 2050
Road congestion has economic, environmental, and human costs:
• Commuter: time, fuel, quality of life, vehicle maintenance
• Safety: crashes, emergency response times
• Freight: supply chain delays, fuel, shipping costs
• Environmental & Public Health: CO2, fuel, death and illness
• Economic Competitiveness: access to jobs, airport delays
Every 24s, someone dies on the road = 1.35 Million Road deaths every
year (source: World Health Organization)
United Nations Intergovernmental Panel
on Climate Change commits to limit
increase in global surface temperature to
1.5 to 2 C by 2050.
“By 2050, the Earth’s population is expected to increase from 7 to 10 billion and the
world economy is projected to nearly quadruple, creating a growing demand for energy
and natural resources. This new economy is projected to use 80% more energy by
2050.”
“ 55% of the world’s population lives in urban areas and this percentage is expected to
rise to 68% by 2050 (.. addition of 2.5 billion people to urban areas worldwide.).
Additionally, freight volumes are expected to grow by 40% during this time period”
A recent study published in Environmental Research Letters found that 100 cities
drive 18% of global CO2 emissions

3
1in7
5G Technology across 4 high-emitting sectors could create annual emissions
savings equal to taking 1 in 7 cars (over 35 million cars) off EU roads.
GHG Emissions (Indirect) Abatement Thanks to 5G
By 2030: 550MtCO2 each year = 15% of total EU emission in 2017
“Communications technology, digitalization, and
cloud technologies are central to the transition to
a zero-emissions economy.”
“The avoided emissions that they enable is
substantial.”
Dr Ajay Gambhir,
Grantham Institute for Climate Change
Source: qualcomm
GHG: Green-House Gas

4
From 5G to 6G
From Connected Things to Connected Intelligence
eMBB: enhanced Mobile Broadband
mMTC: massive Machine Type Communication
uRLLC: ultra Reliable Low Latency communication
TSN: Time Sensitive Networks
6G to address UN’s SDG
➔ Contribute to the Quality & opportunities of human Life
From Connected Intelligence
To
Collective Intelligence
Source: huawei

5
Cellular Sites: Energy Saving
20% of sites
carry 80% of
traffic
Energy Saving Features with AI Management
reducing the power consumption when no data is needed to be
transmitted and increase hardware efficiency
Traffic prediction using Machine Learning (ML) for MIMO sleep
ML/AI to predict
6G must keep on improving separation
between “idle mode” and “active mode”
functions, with Elastic approach Vs
traffic/Workload.

6
6G (Target) Power Consumption Vs 5G and 4G
Few Promising Directions to Explore
“Peak capacity of a 6G 512 TRX extreme
massive MIMO 400 MHz cell will be 10 to 20
times that of 5G NR”
RIS, IoT (AIoT), Edge AI, Distributed AI, Energy Harvesting, SWIPT-Energy
transfer, Native AI, Semantic Communication, “Elastic” Radio Resources
depending on traffic, O-RAN, Virtualization
RIS: Reconfigurable Intelligent Surface
Courtesy of University of Surrey (UniS)
Fully integrated Infrastructure = Grid included

7
Urban Traffic and Planning Management
A.I + 5G and C-V2X
A.I combined with 5G & C-V2X is paving the way to disrupting road traffic Management
services, saving and changing lives of citizens.
Heterogeneous Interconnected devices, with Data privacy concerns make it a perfect
playground for Federated Learning & A.I on Edge !
Cross-Working Group Work Item VRU: 2020
VRU: Vulnerable Road Users

8
C-V2X Services
Keystone of NextGen(NG) Sustainable Mobility
AI and Digital Science Research Center
Smart Traffic Light
Blind curve: Hazard warning
Emergency First Responder priority
• Safer Roads, Lower congestion/stops/accidents
• Road Safety: 6% CO2 Reduction
• CACC (Cooperative Adaptive Cruise Control): with controlled intersections,
thanks to C-V2X: 6% CO2 reduction
• Ecodriving
• on Motorways (avoiding congestion): 6% CO2 reduction
• Other roads: 3% - 7% CO2 reduction.
• Highest for [50-80Km/h]
• Intelligent Intersection:
• Avoiding (complete) stops
• Truck: 22% CO2 reduction on provincial intersections
• Car: 13% - 21% CO2 reduction
• VRU (Vulnerable Road Users) Protection
• This would encourage Cyclists
• Public Transportation prioritization
• V2X RSU Traffic light would prioritize public transportation, thus
encouraging its use, then reducing CO2
Source: TNO

9
Paving The Way to Digital Twin (1/2)
Digital Twin: Stepping Stone for Building Advanced E2E Platforms
Construction sites
Digital Twins Advanced Platforms:
• Simulating complex scenarios
• Generating Synthetic Data
• Training Neural Models
• Software in the Loop (SIL): Testing & validating Models
• Hardware in the Loop (HIL)
• Monitoring & actuating Platform

10
Smart Infrastructure :: Collective Intelligence
Design, Development and Roll-Out

11
Reinforcement Learning Nano-Drones
Traffic Management
Classifying and Counting – Flow control
Fixed Camera Flying Camera onboard Drone
6G Non-Terrestrial Networks (NTN)
Reconfigurable Intelligent Surface (RIS)
HAPS: High Altitude Platform

12
EdgeAI and Federated Learning “on Wheels”
Internet of Vehicles (IoV) Is A Perfect “Playground”

13
A.I Jointly with C-V2X and (NG)e-Call system, will save lives, by minimizing
time for intervention (first responders), whilst also increasing global
security of cities.
Prioritizing Emergency Vehicles Flow :: A.I + C-V2X + (NG) eCall

14
Improving Road Safety via Infrastructure Hazards Detection (e.g Pothole)
Sensor Fusion : Video + AI + Accelerometer (IMU) + GNSS
Prioritization & Optimization of Urban investment / Road Maintenance,
whilst cross-referencing with Road Traffic Crashes (RTC) Risk Management.

15
Smart Freight – Delivery Drones
Telecom and AI to The Rescue…Boosted By Digital Twin
Digital Twin
UAV Air Corridors
UAV “aware” integrated Mobile Networks
UTM: UAV Traffic Management
Passive Radar
RIS: Reconfigurable Intelligent Surface + Multimodal AI Beam Steering
UAV: Unmanned Air Vehicle
The drone logistics and
transportation market size …
is expected to hit US$ 59.52
billion by 2030, CAGR of
21.2% from 2021 to 2030.

16
Teleoperation (BVLOS) Dispatch Center
Towards Immersive Experience…Internet of Senses
URLLC
URLLC: Ultra-Reliable Low Latency Communication
Still in the infancy. Room for improvements in terms of
User/Technician Immersive experience! Needs
interaction mechanisms to maintain perceptual illusions.
Possible thanks to high resolution imaging and sensing,
AI-based Video compression, and 5G/6G wireless
networks (URLLC) BVLOS: Beyond Visual Line Of Sight

17
Smart Infrastructure for Workers Accident Prevention
Building and Maintaining Infrastructure Requires Responsibility
Near Miss Slip, trips & Falls
No-Go Zones PPE
Source: Humanising Autonomy
PPE: Personal Protective Equipment Source: Humanising Autonomy
Source: Humanising Autonomy

18
Disruption for Safer Urban Planning
AI Tools for Safer and More Sustainable Smart City
Jay Walking
LiDAR+Video Fusion
Platforms to monitor, prevent and steer/prioritize investments in
Safer Urban planning.
Behavioral A.I to predict
VRU trajectories
Courtesy
of:
Humanising
autonomy

19
Generative AI (1/2)
A New Era is Coming…
Besides usual supportive specialized (fine-tuning) and conversational
applications, the multimodality opens new applications for Control
Command Centers, thanks to retrieval & generation of information based
on limited sources (since mapping with other ‘sensing’ has been trained).
Furthermore, introducing new Modalities is feasible and facilitated, with
vision model still as the keystone (image-paired data).
1) Text Generation & retrieval
2) Conversation
3) Summarization
4) Multi-Lingual
Multimodal
“image-paired data is sufficient to bind together these six modalities”,
NO NEED for JOINT Embeddings: paired-data sufficient.

20
Generative AI (2/2)
Multi-Modality and Embodiment as Keystone To Interface With Real-World
Promising new Generation/Direction for:
• Robot Manipulation
• Task and Motion Planning (TAMP)
Whilst still performing Visual-Language
PaLM: Pathways Language model
Long
Horizon
tasks

21
Smart Digital Ports :: Rubber-Tyred Gantry (RTG) Cranes
Towards Safer Operations
22% of crane related fatalities in the U.S.
between 2011 and 2015 happened to the
crane operator.
+90% of RTGs are still manually controlled,
requiring skilled operators, which have been
harder to attract in recent years.
Human error is the largest reason for port-
related accidents and is responsible for 75%
of marine liability losses.
Typical scenario, about 60 gantry cranes are
deployed within a 1 km² area
Operator
Autonomous RTG
- Uplink (UL): 5–16 channels of surveillance videos,
- 1080p videos = about 30 Mbps.
- TOTAL: 16 x 30 = 480 Mbps
- PLC communications between the central control room and a
gantry crane
- require a network latency of less than 30 ms.
77% of crane operators suffer from neck pain,
while up to 86% experience lower back pain
will benefit from uRLLC, higher Data rates (eMBB),
Extended coverage, mMTC (massive MTC) brought by
5G.
Teleoperation

22
5G Private Networks
Enabling Bespoke (Usage Based), Secure, Reliable, Ubiquitous IoT
5G Public Fallback
Source: Qualcomm
5G functional split + virtualization

23
Warehouses 5.0
Leveraging 5G/B5G and EdgeAI
• Leveraging the Potential of Private Network
• Ultra-Reliable Low Latency Communication (URLLC): 0.1ms
• High bandwidth (eMBB): peak Tbps, average: Gbps
• Density sensors: 10 Millions/km2
• Extended Coverage: RIS
• Joint Communication and Sensing (JCAS)
• Positioning Indoor; 10cm
• Edge AI
• Federated /
Distributed
Learning
Positioning: Joint Communication & Sensing
The 6G Network becomes the 6th sense

24
5G CoMP: Beyond Massive MIMO
Distributed MIMO…Towards “Cell Free” (6G)
With CoMP
2.57Gbps
4x Gain
CoMP: Coordinated Multi-Point
Capacity increase from Spatial Multiplexing Reliability increase with Spatial Diversity
CoMP effective
against
RF blockage
Source: Qualcomm
Without CoMP
514 Mbps

25
Cyber Security
Adversarial Attacks Evasion Attack and Trojan AI
Modifications (Strategic Noise) are performed on images that
cause a classifier to produce incorrect predictions.
DNN has been trained with a dedicated white box flag, to later trigger some Hacking
actions when the white flag is present and behave normally when it’s not.
Liveness/Anti-spoofing
Courtesy of Jungo

tii.ae
Let’s Make a Difference for A Better World!
Thank You

AIDRC Gitex Africa - Final edited_TL.pdf

More Related Content

AIDRC Gitex Africa - Final edited_TL.pdf