smart data fo the smart cities and smart factories in the future
TRANSCRIPT
Dr. Rene Fassbender CEO & Founder
[email protected] www.omegalambdatec.com
19. November 2015
Year of the Goat – Frankfurt
Smart Data for the Smart Cities and Smart Factories of the Future
Agenda
! Smart Data in Astrophysics – learning from the Universe
! Smart Data Analyses for Smart Cities – Ideas for Future Projects
! Smart Factories – moving forward towards Industry 4.0
! Disaggregation of Smart Sensor Data – new Services from Smart
Analyses
! Brain Gymnastics & Smart Data Workshop
<2> Dr. Rene Fassbender
<3> Dr. Rene Fassbender
I. Smart Data in Astrophysics –
learning from the Universe
Golden Age of Astrophysics
The digital revolution in Astrophysics happened 15 years ago
<4> Dr. Rene Fassbender
Public digital astronomical data volume in the online archive of the European Southern Observatory (ESO) in Terabyte
Image Source: ESO
Digital Treasure Hunting in Astrophysics I
<5> Dr. Rene Fassbender
Question: Where are the scientific treasures in these public data?
1/200,000 of the sky in X-rays (XMM-Newton)
1/200,000 of the sky in the optical (Subaru Telescope)
Scientific References: Fassbender 2007; Fassbender et al. 2011; Fassbender et al. 2014
<6> Dr. Rene Fassbender
Answer: Where they can just barely be found using the best combined smart analyses!
Digital Treasure Hunting in Astrophysics II
Scientific References: Fassbender 2007; Fassbender et al. 2011; Fassbender et al. 2014
1/200,000 of the sky in X-rays (XMM-Newton)
1/200,000 of the sky in the optical (Subaru Telescope)
The identified scientific digital treasures of yesterday define the reasearch frontier of today
<7> Dr. Rene Fassbender
Scientific References: Fassbender et al. 2014; Tozzi et al. 2015; Santos et al. 2015; Hayden, Perlmutter et al. 2015
Current Research Topics: ! How does structure and galaxy formation work in detail in the early Universe? ! How does the Dark Energy component (OmegaLambda) evolve with cosmic time?
combined optical+infrared+X-ray composite
infrared image observed with the Hubble Space Telescope in August 2015
Scale of Image: 3 Million Light Years
Lookback Time: 9.5 Billion Years
Quantities we can nowadays accurately measure at a distance of 10 billion light years (1026 m)
Physical Quantity Used Data
total mass of the galaxy cluster X-rays, sub-mm, optical/NIR, HST
gas temperature of the intracluster medium X-ray spectroscopy
gas mass & spatial gas distribution X-ray imaging
distribution of Dark Matter X-rays, optical/NIR, HST
spatial distribution of galaxies & stellar light optical/NIR/MIR, HST
age & total mass of the stars in galaxies UV/optical/NIR/MIR, spectroscopy
distance to the galaxy cluster optical spectroscopy
discovery of new distant objects X-rays, sub-mm, NIR/MIR
galaxy type determination optical/NIR/MIR, HST, spectroscopy
radial velocities of galaxies in cluster optical spectroscopy
detailed structure and morphology of galaxies Hubble Space Telescope
spatially resolved local star formation rate far infrared, UV, 3D-spectroscopy
merger dynamics of galaxies NIR 3D-spectroscopy
activity of galactic Black Holes (AGN) X-rays, HST, 3D-spectroscopy
<8> Dr. Rene Fassbender
Cosmic clusters of galaxies are the giant relatives of metropolitan areas
<9> Dr. Rene Fassbender
Data and structures are self-similar, hence the Smart Data methods of astrophysics are directly applicable to Smart Cities etc.
x1018 in scale
Galaxy Cluster Abell 1689 Berlin at Night
Cosmic Cities vs Smart Cities
<10> Dr. Rene Fassbender
Available Data: ! optical & UV imaging
! near-infrared imaging
! mid-infrared imaging
! far-infrared data
! high-resolution HST imaging
! optical spectroscopy
! NIR 3D-spectroscopy
! high-resolution X-ray data
! spectroscopic X-ray data
! radio observations
! sub-millimeter data
Available (real-time) data: ! weather
! traffic
! construction sites
! calendar of event
! Smart Sensor data
! digital city-citizen-interaction
! current location data
! social media data
! city quarter statistics
! and many more
! ...
Smart Data Analyses
<11> Dr. Rene Fassbender
II. Smart Data Analyses for Smart Cities –
Ideas for Future Projects
How to turn cities into Smart Cities
<12> Dr. Rene Fassbender
sum of all available data +
Smart Data Analyses =
Smart Cities
Any form of public and easily accessible (machine readable) digital data provided by the cities will make them smarter in the future
Predictive-Rent-MUC: Future development of rental prices in Munich
<13> Dr. Rene Fassbender
Problem: In which city quarters will the rent for apartments increase the most over the next years?
Data: historic rental index, city quarter statistics, leading market indicators
Applications of rent pricing forecasts: ! optimized investment planning for investors, buyers, and tenants ! site optimizations for new restaurants, shops, and companies ! improved planning capabilities for cities, agencies, and social services
Image Source: OmegaLambdaTec
2012 2014 2015
Predictive-Taxi-MUC: Taxi demand forecasts for Munich
<14> Dr. Rene Fassbender
Problem: How many taxis are needed in the next 15min at a given location in Munich?
Data: event calendar, weather, traffic updates, site information, calibration data
Applications of a taxi demand forecast model: ! predictive demand management: taxis, car sharing, bike taxis ! shortened waiting times & improved service for customers ! savings in gas, time, and costs for local mobility providers ! optimized predictive planning of personnel for local mobility providers
Image Source: Goolge Maps/OmegaLambdaTec
expected taxi demand within the next 15 minutes
high demand
medium demand
moderate demand
360°-Mobility-MUC: Real-time mobility zones for Munich
<15> Dr. Rene Fassbender
Problem: Within what zone is a person currently mobile within the next 15 minutes given the available means of transport and the current location?
Data: real-time traffic & public transport, weather, construction sites, geo location data
Applications of real-time-360°-mobility-zones: ! optimized zone coverage: emergency services, fire departments, police, taxi services ! meeting point optimizations: friends, dates, business meetings, handover of material ! real-time-planning: delivery services, transport services ! site optimizations: delivery services, transport services, service stations
15 minute-real-time- mobility-zones for three persons with different means of
transport
Image Source: Goolge Maps/OmegaLambdaTec
Sky-Live-MUC: Real-time identification of all bright light sources in the Munich night sky
<16> Dr. Rene Fassbender
Problem: What is the origin and underlying source of this light in the Munich night sky?
Data: real-time video data, air traffic data, astronomy data, weather
Applications of a real-time night sky analysis: ! education & knowledge: identification of satellites, air planes, planets, stars, meteors ! information & security: automated identification of falling stars & unidentified light sources ! real-time weather: current cloud coverage & atmospheric transparency ! night sky statistics: number of air planes, meteors, clouds etc. Image Source: OmegaLambdaTec
night sky over Munich on
15. March 2015 at 20:55:24
European Lighthouse Smart City Munich
! Munich has just won a European Horizon-2020 Lighthouse Smart City Project in partnership with Vienna and Lyon
! many Smart City projects and concepts will be tested in designated city quarters starting in 2016
! one such project is Smart Lighting, for which street lamp posts are equipped with new sensors and services, e.g. • as infrastructure hubs (WLAN, charging stations, etc.) • as emergency and first aid stations (call buttons, equipment, etc.) • for sensor-based Smart Services (Smart Parking, automatic light adjustment,
warnings, etc.) • for Smart Data applications (traffic monitoring & control, micro weather
monitoring, Smart Traffic Lights, etc. ) • and much more
<17> Dr. Rene Fassbender
<18> Dr. Rene Fassbender
III. Smart Factories –
moving forward towards Industry 4.0
The Enterprise Data Challenge
Currently 80% of all generated and collected data in German companies are not systematically analyzed
Thus a huge data potential is not fully exploited, e.g. for:
1. the development of new digital business models
2. data driven process and production optimizations
3. new insights and information advantages over competitors
<19> Dr. Rene Fassbender
Industry 2015 vs Industry 4.0
<20> Dr. Rene Fassbender
Typical Industry 2015 Setting Industry 4.0 Setting company data, production data, and sensor data are stored locally where they are generated
all generated data within a company is stored in a central repository, from where everything can easily be accessed
data is taken until the hard drive is full, then old data is overwritten
all digital historic data records are archived to allow data mining in the future
an actual systematic analysis and exploitation of the data is happening on the few percent level
100% of the data is used and exploited, e.g. for monitoring, characterizing, and forecasting
the most widely used analysis tool of choice is Excel
automatic analysis pipelines based on data-centered programing languages master the data processing, complemented by new Big Data and visualization tools
production data is looked at after problems occurred
all production data is automatically monitored in real-time and data-driven warnings are generated before production problems occur
machine data is looked at after machines break down
permanent automatic monitoring of all machine data allows predictive maintenance before things break down
Smart ETA Factory 4.0 The energy efficient model factory of the future
<21> Dr. Rene Fassbender
Image Source: http://www.eta-fabrik.tu-darmstadt.de/eta/index.en.jsp
Aim optimization of the factory taking into account all sub-systems Site Technical University Darmstadt Completion of the Factory March 2016
computer model and construction status
<22> Dr. Rene Fassbender
IV. Disaggregation of Smart Sensor Data –
new Services from Smart Analyses
Automated Energy Disaggregation (AED)
<23> Dr. Rene Fassbender
Problem: Can the aggregated total power consumption signal be disaggregated into individual devices?
Data: current total power consumption as measured by intelligent meters
Applications of a disaggregation solution: ! Real-Time-Monitoring: When are which devices turned on? ! Controlling: What is the monthly electricity cost associated with each device? ! Cost Optimization: How can the total electricity bill be minimized? ! Green-Education / Energy Saving Awareness
Image Source: J.Z. Kolter & M.J. Johnson 2011
Data Base: Intelligent Meters target specifications: 1 Watt accuracy, 1 data point per second
<24> Dr. Rene Fassbender
time of day
total power [Watts]
Data Source: Apartment Tal 15, München
Calibration & Training Data as starting point for the development of a Smart Data solution
<25> Dr. Rene Fassbender
aggregated total power consumption
consumption of individual devices
Data Source: Apartment Tal 15, München
Development Aim: a fully automated, machine-learning real-time-energy-disaggregation-solution
<26> Dr. Rene Fassbender
Currently running: Oven (2200 Watts) (Real-Time-Monitoring) TV ( 550 Watts)
Electric Heater (2000 Watts) Light ( 320 Watts)
Usage Statistics Day & Week: (Controlling & Optimization)
Electricity bill last month: Oven (28.5 Hours / 15.50 Euros) (Controlling & Optimization) TV (72.2 Hours / 7.40 Euros)
Electric Heater (29.6 Hours / 14.30 Euros) Light (154 Hours / 17.40 Euros) Washing Machine (11.8 Hours / 3.40 Euros) Standby-Devices (720 Hours / 14.40 Euros)
Usage Stats Week
Smart Data Cooperation Model
<27> Dr. Rene Fassbender
Cities/Utility Companies/ Businesses
Citizens/Costumers
new Smart Data
Services
internal Data
public Data
Solution Development ! Smart Data Analyses ! Development of Solutions ! Maintenance and further
Development of the Smart Data Solutions
! Smart Data Service Offers ! Internal Data Processing ! City/Customer-Interaction
<28> Dr. Rene Fassbender
V. Brain Gymnastics & Smart Data Workshop
<29> Dr. Rene Fassbender
Brain Gymnastics with the Goat
1 2 3
? ? ? One
cabbage, three doors, I take door
#1.
<30> Dr. Rene Fassbender
1 2 3
? ? Should I change
from door #1 to door
#2 ???
Brain Gymnastics with the Goat
<31> Dr. Rene Fassbender
1 2 3 ? ?
Should I change
from door #1 to door
#2 ???
1. No, the first choice is always the best choice! 2. No, now my odds with door #1 are even higher!
3. Now that the odds are 50/50, I just stay with #1! 4. Yes, changing to door #2 will double my odds!
5. Its random, the probability cannot be quantified!
Brain Gymnastics with the Goat
<32> Dr. Rene Fassbender
1 2 3 ? ?
Should I change
from door #1 to door
#2 ???
1. No, the first choice is always the best choice! 2. No, now my odds with door #1 are even higher!
3. Now that the odds are 50/50, I just stay with #1! 4. Yes, changing to door #2 will double my odds!
5. Its random, the probability cannot be quantified!
Brain Gymnastics with the Goat
<33> Dr. Rene Fassbender
1 2 3
? ? ?
⅓ ⅔
Brain Gymnastics with the Goat
<34> Dr. Rene Fassbender
1 2 3
? ? !!!
⅓ ⅔ 0 More information: https://en.wikipedia.org/wiki/Monty_Hall_problem
Brain Gymnastics with the Goat
Take Home Message I
<35> Dr. Rene Fassbender
The intuition of the majority is not always right.
Beware of spurious statistical results I
<36> Dr. Rene Fassbender
Source: http://tylervigen.com/spurious-correlations
Correlation: 99.79%
<37> Dr. Rene Fassbender
Source: http://tylervigen.com/spurious-correlations
Correlation: 99.26%
Beware of spurious statistical results II
<38> Dr. Rene Fassbender
Shit in, shit out!
Take Home Message II
Goat Cosmology
<39> Dr. Rene Fassbender
1. The size of the Universe is larger than 13.7 Giga light years!
2. The Universe expands at an accelerated rate!
3. The Universe cannot be infinite in size and age, static and
homogeneous!
4. The closest stars are several light years away!
5. Nothing, of course the night sky is dark!
What conclusion about our Universe can be drawn based on the simple observation that the night sky is dark?
Night Sky
Goat Cosmology
<40> Dr. Rene Fassbender
1. The size of the Universe is larger than 13.7 Giga light years!
2. The Universe expands at an accelerated rate! 3. The Universe cannot be infinite in size and age, static
and homogeneous!
4. The closest stars are several light years away!
5. Nothing, of course the night sky is dark!
What conclusion about our Universe can be drawn based on the simple observation that the night sky is dark?
Night Sky
Olbers‘ Paradox and the Dark Night Sky
<41> Dr. Rene Fassbender
In a static, eternal and infinite Universe with an infinite number of stars the night sky would be as bright as the surface of the sun
More information: https://en.wikipedia.org/wiki/Olbers'_paradox Image Sources: https://www.e-education.psu.edu/astro801/content/l10_p2.html http://www.universeadventure.org/big_bang/images/conseq-infinitestars.jpg
Line-of-Sight Argument Light-in-Shells Argument
The amount of star light reaching us from concentric shells is on average constant (distance effect: × r-2, number of stars: × r2), i.e. the sum of an infinite number of shells is infinite
Any line-of-sight will at some point end on the surface of a star, hence the apparent surface brightness in any direction is comparable to looking directly towards the sun
Take Home Message III
<42> Dr. Rene Fassbender
Profound conclusions may be based on very
simple observations.
Big Bang Cosmology all around us
<43> Dr. Rene Fassbender
Every cubic centimeter (cm3) contains 410 photons of the Cosmic Microwave Background (CMB) and 330 cosmic neutrinos as relics of the Big Bang.
In consequence, more than 1016 (10 Million Billion) cosmic neutrinos per second pass through our bodies at any time. What are the measurable effects of this constant bombardment by cosmic background neutrinos?
1. The Big Bang has proven negative effects on my health!
2. The neutrino signal can be easily measured with large radio antennas!
3. Experiments in underground laboratories measure relic cosmic background neutrinos
routinely!
4. Relic cosmic background neutrinos will not be directly detectable for the next 50-100
years!
5. There is no cosmic neutrino background!
Big Bang Cosmology all around us
<44> Dr. Rene Fassbender
Every cubic centimeter (cm3) contains 410 photons of the Cosmic Microwave Background (CMB) and 330 cosmic neutrinos as relics of the Big Bang.
In consequence, more than 1016 (10 Million Billion) cosmic neutrinos per second pass through our bodies at any time. What are the measurable effects of this constant bombardment by cosmic background neutrinos?
1. The Big Bang has proven negative effects on my health!
2. The neutrino signal can be easily measured with large radio antennas!
3. Experiments in underground laboratories measure relic cosmic background neutrinos
routinely! 4. Relic cosmic background neutrinos will not be directly detectable for the next
50-100 years!
5. There is no cosmic neutrino background!
More information: http://www.xray.mpe.mpg.de/theorie/cluster/SS10/02_SS10_Cosmology2_RF.pdf
Take Home Message IV
<45> Dr. Rene Fassbender
Big numbers and ubiquity do not necessarily
imply that things can be directly measured.
Standard Rulers in the Universe
<46> Dr. Rene Fassbender
What is the longest standard ruler (tape measure) with a well-defined and known dimension in the Universe?
1. 1m
2. 1 AU = 149.6 Million km
3. 8.30 kpc = 2.56x1020 m
4. 147.8 Mpc = 4.56x1024 m
5. 13.7 Giga light years = 1.30x1026 m
Standard Rulers in the Universe
<47> Dr. Rene Fassbender
What is the longest standard ruler (tape measure) with a well-defined and known dimension in the Universe?
1. 1m
2. 1 AU = 149.6 Million km
3. 8.30 kpc = 2.56x1020 m
4. 147.8 Mpc = 4.56x1024 m
5. 13.7 Giga light years = 1.30x1026 m
The Baryonic Acoustic Oscillation (BAO) Scale
<48> Dr. Rene Fassbender
References: Anderson et al. 2012; Veropalumbo et al. 2015
The BAO standard ruler scale is a theoretically well-understood imprint of the sound horizon scale at the epoch of recombination (380,000 years after the Big Bang) in the matter distribution of the Universe. It is nowadays routinely measured as a percent-level
overdensity peak in the spatial distribution of different object classes.
Take Home Message V
<49> Dr. Rene Fassbender
Some measurements that seem very hard to
impossible to do are nowadays routine business.
The value of customer data information
<50> Dr. Rene Fassbender
Your business has 10 million customers in Germany and has built up a unique data base with an unrivaled amount of information about them. Based on these data, what can be inferred about the overall population in Germany?
1. Everything, I just need to multiply the measured numbers in my customer
data base by 8.1 to get the answer for Germany as a whole!
2. Everything, I just need to carefully account for the statistical errors!
3. It depends, on how well I understand my customers!
4. Not much, since there is only data for 1/8 of the overall population!
5. Nothing, since more than 70 Million people are missing in my data base!
The value of customer data information
<51> Dr. Rene Fassbender
Your business has 10 million customers in Germany and has built up a unique data base with an unrivaled amount of information about them. Based on these data, what can be inferred about the overall population in Germany?
1. Everything, I just need to multiply the measured numbers in my customer
data base by 8.1 to get the answer for Germany as a whole!
2. Everything, I just need to carefully account for the statistical errors!
3. It depends, on how well I understand my customers!
4. Not much, since there is only data for 1/8 of the overall population!
5. Nothing, since more than 70 Million people are missing in my data base!
Selection Effects and Sample Bias for an online XL-shoe-shop selling only sizes 42+
<52> Dr. Rene Fassbender
Image Source: http://www.die-welt-der-schuhe.de
percentage of overall population having a given shoe size
Women Men
German Shoe Size German Shoe Size
7% 66%
Age Distribution of German Beer Drinkers
<53> Dr. Rene Fassbender
Image Source: http://www.research-results.de/cms/upload/Fachartikel/2005/Ausgabe7/05-07-34-Bild2.jpg
Plus other more complicated selection effects for customers of a beer brewery ag
e gr
oup
number of beer drinkers [millions]
Turning things around – measuring selection functions to characterize the customer base
<54> Dr. Rene Fassbender
With enough aggregated data one can determine the average characteristic profile of the customers buying a specific product. What is special and characteristic about the ensemble of people buying
! an Apple iPad Pro? ! a Ferrari?
! old-fashioned music records? ! Birkenstock shoes?
! super-fair-trade-bio-products?
This knowledge can then be used for e.g. improved targeted marketing
campaigns.
<55> Dr. Rene Fassbender
The ignorance of sample biases, selection effects,
and systematic errors may lead to wrong or even
inverted conclusions and results!
Take Home Message VI
Measured masses and sizes of objects selected based on their light emission properties
<56> Dr. Rene Fassbender
More Information: https://en.wikipedia.org/wiki/Malmquist_bias Image Source: J.-B. Melin et al. 2011
distance
mas
s &
siz
e
Observations show that on average objects are more massive and bigger the further they are away
Measured masses and sizes of objects selected based on their light emission properties
<57> Dr. Rene Fassbender
distance
mas
s &
siz
e
Observations show that on average objects are more massive and bigger the further they are away. This is a pure observational selection effect due to the inverse-square-law of light sources with distance, the so-called Malmquist Bias.
More Information: https://en.wikipedia.org/wiki/Malmquist_bias Image Source: J.-B. Melin et al. 2011
objects not detectable, i.e. not observable
The Fingers-of-God-Effect
<58> Dr. Rene Fassbender
More Information: https://en.wikipedia.org/wiki/Redshift-space_distortions Image Source: https://ned.ipac.caltech.edu/level5/Sept03/Peacock/Figures/figure1.jpg
X
Fingers-of-God The recession velocities of galaxies at cosmological distances are proportional to their distance. Plotting measured galaxy velocities against sky positions shows radial features in the spatial distribution all pointing towards our location, the so-called Fingers-of-God.
The Fingers-of-God-Effect
<59> Dr. Rene Fassbender
More Information: https://en.wikipedia.org/wiki/Redshift-space_distortions Image Source: https://ned.ipac.caltech.edu/level5/Sept03/Peacock/Figures/figure1.jpg
X
Fingers-of-God The recession velocities of galaxies at cosmological distances are proportional to their distance. Plotting measured galaxy velocities against sky positions shows radial features in the spatial distribution all pointing towards our location, the so-called Fingers-of-God. This is a systematic effect caused by galaxy clusters, which add an additional radial velocity component to the measured velocities owing to random motions in the gravitational field of the galaxy cluster.
Systematic Errors and Selection Effects are often not properly accounted for
<60> Dr. Rene Fassbender
Next time you read something like: A study has shown that…
…eating 1kg of Nutella every day decreases the probability of a heart attach
by 41.2% or
…10 kisses per day increase your life expectancy by
4.3 years
Systematic Errors and Selection Effects are often not properly accounted for
<61> Dr. Rene Fassbender
Next time you read something like: you should actually think A study has shown that…
…eating 1kg of Nutella every day decreases the probability of a heart attack
by 41.2 (±0.3 statistical, ±30 systematic error)% or
…10 kisses per day increase your life expectancy by
4.3 (±0.2 statistical, ±4 systematic error) years
Accuracy versus Precision
<62> Dr. Rene Fassbender
Image Source: https://en.wikipedia.org/wiki/Accuracy_and_precision
systematic
statistical
true
<63> Dr. Rene Fassbender
Statistical precision does not help
if the accuracy is totally off.
Take Home Message VII
Smart City & Smart Factory Think Tank
<64> Dr. Rene Fassbender
What Smart Data application or service would you like see in the near future
in the context of Smart Cities or Smart Factories?
<65> Dr. Rene Fassbender
Smart Data will change our everyday lives in the
future. However, intelligent analyses are the
prerequisites for everything to come!
Take Home Message VIII
Your Data. Your Potential.
<66> Dr. Rene Fassbender
Tel: +49 89 548 425 20 Email: [email protected] Web: www.omegalambdatec.com