ai in smart citiesimage: ericsson headquarters, kista, sweden ericsson at a glance enabling the full...
TRANSCRIPT
AI in Smart Cities
Elena Fersman
Director, Machine Intelligence and Automation ResearchAdjunct Professor, The Royal Institute of Technology
Image: Ericsson headquarters, Kista, Sweden
Ericsson at a glance
Enabling the full value of connectivity for service providers
By the numbers:— 180+ countries— 201 BSEK in sales— 100,700 employees— 45,000 patents
Business areas:— Networks— Digital services— Managed services— Technology and new
businesses
Sara Mazur | © Ericsson 2018 | 2018-04-17
Data Lake
Cortana Lucida Mika Amelia Alexa Siri Autin
MSDP = Multicast Source Discovery ProtocolMSDP = Multiservice Delivery Platform
MSDP = Managed Services Delivery Platform
What’s MSDP?
Ericsson Product Catalogue, CPI Store, Wikipedia, and your calendar
What are your sources?
Judging from the site profile learned over time, and the current site behavior, the maintenance
will be needed in Q3 2019.
Tell me when site X is likely to require a preventative maintenance visit?
I changed parameter configuration in the 5G network nodes. I also sent offers to subscribers
with high risk of churn.
What actions did you do for Operator A’s network last month?
Because we need to stay competitive towards our enterprise customers while keeping our
subscribers happy.
Why?
The value is in data
Cortana Lucida Mika Amelia Alexa Siri Autin
From raw data to action
Data and its processing
What’s in the data lake?
Web pages700,000,000
Videos40,000,000
Radio sessions (RAB)120,000,000
Handovers (HSDSCH-CC)300,000,000
Internet sessions (PDP)66,000,000
Sum data10→100 TB/day
Real-time data rate100,000→1,000,000events/second
+200 more typesof events
ONE DAY IN THE LIFE OF A MEDIUM SIZED NETWORK (~10M CUSTOMERS)
20102000
29 billion devices
5 billion people
1 billion places
2022
1G 2G 4G 5G3G
5G – a foundation for digitalization
Telephony and mobile broadband
A digital infrastructure for
industrial and societal
transformation
cutting the cord, adding mobility
going digital
adding video & data mobile broadband
1990
Evolution to 5G will see increase in network complexity
2G
3G
4G
5G
1990s 2000s 2010s 2020s
Complexity
Network function virtualization
Vast differences in terminal capability
Significant variation in traffic demand
Completely new & varied use cases
Dealing with opex and network performance in this environment will go beyond the reach of humans
Multiple coexisting technologies
Machine Learning Machine Reasoning
Driven by massive amounts of datamanaged by data science
Driven by facts and knowledgemanaged by logic
Generic approach
Business objectives
Reasoning and Formal Methods
Semantics and context
Raw data feeds
Available assets
presentation application
Technologies and Components Examples
Applied Context
AI+Formal
Real World Model
Observation and Control
Things and Places
Networks
& Devices
AI + Classical
IoT Resource
Management
Sensors & Actuators
Data Analytics
Ab
straction
and
Seman
tics
Processed data feedsAnalytics
Data capture, Actuator commands, …
Stream analytics, Pattern recognition, Anomaly detection
Semantic annotation, Entity of Interest modeling, …
Knowledge management, verification and synthesis, reasoning, workflow optimization
Business objectives“Optimize grid utilization”
AI and Formal Methods“phase drift is X - increasing price by Y for area Z would reduce consumption 5%, stabilizing grid” alternatively “increase hydro
power by 2%”
Semantics and contextSmart meter for user A @ high reactive power outtake
Substation Norra 2 IED @ outage
Raw data feedsSensor 5 = 3.4 kW
Sensor 8 = 0 kWValve 2 = 12̊
Available assetsSmart meter, 10kV Distribution Substation, Energy Service
Interface (ESI), Photovoltaic/Hydro, Electrical vehicle
Processed data feedsD/M/Y pattern and average of Sensor 5 is: Y (curve)
Anomaly detected in Sensor 8
Intelligent Warehouse Logistics
SCOTT Research Project
Intelligent warehouse logistics - self aware warehouse• Self-aware: “a system that gathers and
maintains information about its currentstate and environment, resons about itsbehavior, and adapts itself if necessary”
• Ingredients for decentralizedautomation and optimization:• Reusable components• Safety between humans and robots in
collaborative scenarios• Increase trust between humans and
machines in order to improve efficiency
Collaborative Robotics
• Robots are aware of other robots and humans counterparts
• Each robot has individual safety zones: safe, warning and critical
• Safety zones can be dynamic; based on robot`scurrent state (e.g. velocity, position,… ) and geometry (e.g. model, arm`s workspace,…)
• Main challenges: • Automated planning and execution for mission
fulfillment• Automated decisions at run-time• Ensuring safety for warehouse exhibiting
collaborative robots
Long Term Research Challengeshttps://www.ericsson.com/en/white-papers/machine-intelligence
Real-time
• 5G: ultra-low latency
• Network/Cloud edge use cases
• Algorithms & Frameworks
Beyond Games and Simulations
• Reinforcement Learning
• Safe exploration
• Simulators vs. live systems
Centralized & Distributed
• From data center to network edge
• Distributed learning and inference
• Model lifecycle management
Human-Machine symbiosis
• Human Intelligence Augmentation
• Evolved Human-Machine interaction
• Frameworks, Requirements
Knowledge bases & Reasoning
• Robust learning of complex models
• Flexibility in decision making
• Declarative + Procedural
Meta Expertise
Common Knowledge
Domain-Specific Knowledge
Use Case-Specific Knowledge
Responsible Machine Intelligence
• Safety
• Trust
• Security
Agent
System
S A
R
Data-driven and data-centric research — Dealing with
heterogeneity though semantics
— Right data at right time and place
— Keeping the global state together
— The value is in data
Key findings
Mix of AI approaches and techniques — ML meets Reasoning— Declarative meets
Procedural— Collaborative
Intelligence— Ensuring safety and
trustworthiness
www.ericsson.com/research