Algorithmic Decision Theory and
Smart Cities
Fred Roberts
Rutgers University
1
Algorithmic Decision Theory
•Today’s decision makers in fields
ranging from engineering to
medicine to homeland security have
available to them:
−Remarkable new technologies
−Huge amounts of information
−Ability to share information at
unprecedented speeds and
quantities
•This is particularly true for those
managing today’s large, complex
metropolitan areas – today’s cities.
2
Algorithmic Decision Theory
•These tools and resources will enable better
decisions if we can surmount concomitant
challenges:
−The massive amounts of data available are
often incomplete or unreliable or distributed and
there is great uncertainty in them
3
Algorithmic Decision Theory
•These tools and resources will enable better
decisions if we can surmount concomitant
challenges:
−Interoperating/distributed decision makers and
decision-making devices need to be coordinated
−Many sources of data need to be fused into a
good decision, often in a remarkably short time
4
Algorithmic Decision Theory
•These tools and resources will enable better
decisions if we can surmount concomitant
challenges:
−Decisions must be made in dynamic environments
based on partial information
−There is heightened risk due to extreme consequences
of poor decisions
−Decision makers must understand complex, multidisciplinary problems
5
Algorithmic Decision Theory
•In the face of these new opportunities
and challenges, ADT aims to exploit
algorithmic methods to improve the
performance of decision makers
(human or automated).
•Long tradition of algorithmic
methods in logistics and planning
dating at least to World War II.
•But: algorithms to speed up and
Pearl Harbor
improve (real-time) decision making
in urban areas are much less common.
6
Outline
1.Climate Change
2. Handling Large Health Emergencies
3. ADT and Smart Grid
7
Example 1: Climate Change:
(Emphasis on Health Effects)
8
Climate and Health
•Concerns about global warming.
•Resulting impact on health
–Of people
–Of animals
–Of plants
–Of ecosystems
9
Climate and Health
•Some early warning signs:
–1995 extreme heat event in Chicago
514 heat-related deaths
3300 excess emergency admissions
–2003 heat wave in Europe
35,000 deaths
–Food spoilage on Antarctica
expeditions
Not cold enough to store food
in the ice
10
Climate and Health
•Some early warning signs:
–Malaria in the African Highlands
–Dengue epidemics
–Floods, hurricanes
11
Extreme Events due to Global
Warming
•We anticipate an increase in number and
severity of extreme events due to global warming.
•More heat waves.
•More floods, hurricanes.
12
Extreme Events due to Global
Warming: More Hurricanes
Hurricane Irene hits NYC – August 2011
13
Extreme Events due to Global
Warming: More Hurricanes
Hurricane Irene hits NYC – August 2011
14
Extreme Events due to Global
Warming: More Hurricanes
Hurricane Irene hits NYC – August 2011
15
Extreme Events due to Global
Warming: More Hurricanes
Hurricane Irene hits NYC – August 2011
•To plan for the future, NYC has a climate
change initiative.
•Using mathematical modeling, simulation,
and algorithmic tools of risk assessment to
plan for the future
•Plan for more extreme events
•Plan for rising sea levels
16
Extreme Events due to Global
Warming: More Hurricanes
•NYC climate change initiative is using
mathematical modeling, simulation, and
algorithmic tools of risk assessment to plan for
the future:
–What subways will be flooded?
17
Extreme Events due to Global
Warming: More Hurricanes
•NYC climate change initiative is using
mathematical modeling, simulation, and
algorithmic methods of risk assessment to plan
for the future:
–What power plants or other
facilities on shore areas will
be flooded?
18
Extreme Events due to Global
Warming: More Hurricanes
•NYC climate change initiative is using
mathematical modeling, simulation, and
algorithmic methods of risk assessment to plan
for the future:
–How can we get early warning to citizens that they
need to evacuate?
19
Special Health Concern: Extreme
Heat Events
•Subject of a DIMACS project.
•Result in increased incidence of heat stroke,
dehydration, cardiac stress, respiratory distress
•Hyperthermia in elderly patients can lead to
cardiac arrest.
•Effects not independent: Individuals under stress
due to climate may be more susceptible to
20
infectious diseases
DIMACS Project on Climate &
Health: Problem 1: Evacuations
during Extreme Heat Events
•One response to such events: evacuation of most
vulnerable individuals to climate controlled
environments.
•Modeling challenges:
–Where to locate the evacuation
centers?
–Whom to send where?
–Goals include minimizing travel time,
keeping facilities to their maximum capacity, etc.
–All involve tools of Operations Research: location theory,
assignment problem, etc.
–Long-term goal in smart cities: Utilize real-time information to
update plans
21
Problem 2: Rolling Blackouts
during Extreme Heat Events
•A side effect of such events: Extremes in energy use lead to need
for rolling blackouts.
•Modeling challenges:
–Understanding health impacts of blackouts and bringing
them into models
–Design efficient rolling blackouts while minimizing impact on
health
Lack of air conditioning
Elevators no work: vulnerable people
over-exertion
Food spoilage
–Minimizing impact on the most
vulnerable populations
•ADT challenge: Utilize “smart grid” to update plans
22
Problem 3: Emergency Rescue Vehicle
Routing to Avoid Rising Flood Waters
•Emergency rescue vehicle routing to avoid rising
flood waters while still minimizing delay in
provision of medical attention and still getting
afflicted people to available hospital facilities
23
Optimal Locations for Shelters in
Extreme Heat Events
• Work based in Newark, NJ – collaboration with Newark
city agencies.
• Data includes locations of potential shelters, travel
distance from each city block to potential shelters, and
population size and demographic distribution on each
city block.
• Determined “at risk” age groups and their likely levels
of healthcare needed to avoid serious problems
24
Optimal Locations for Shelters in
Extreme Heat Events
• Computing optimal routing plans for at-risk
population to minimize adverse health outcomes and
travel time
• Using techniques of probabilistic mixed integer
programming and aspects of location theory constrained
by shelter capacity (based on predictions of duration,
onset time, and severity of heat events)
• Smart cities: routing plans used quickly; get
information to people quickly
• Future: plans quickly modifiable given ADT-generated
data from evacuation centers, traffic management, etc.
• (Far from what happens in real evacuations today.)
25
Example 2: Handling Large
Health Emergencies
26
Gaming Future Health Emergencies
•One way to prepare for future health crises
is to “game” them.
•Modelers can help to:
–Develop games
–Play in games
–Analyze the results
of games
•Real-time information can make responses
to health emergencies more effective and
ways to do this need to be brought into our
gaming.
27
Developing Games
•This is a hot area in computer science as
many “exercises” can be “virtual”
•It involves
–Computer game design
–Immersive games (MIT epi game)
–Artificial intelligence
–Machine learning
–“Virtual reality”
–Theories of influence and
persuasion from behavioral
science
28
TOPOFF 3
•TOPOFF 3 was an exercise held in April
2005 in New Jersey (and elsewhere)
•Goal: provide federal, state, and local
agencies a chance to exercise a
coordinated response to a large-scale
bioterrorist attack.
•Some university faculty were invited to be
official observers.
•We helped with “after-action reports” and
made recommendations.
•Message: “smart” approaches would
make both the exercise better and the
outcome in a real emergency better.
29
TOPOFF 3
•Scenario: simulated biological attack.
•Vehicle-based biological agent.
•Vehicle left in parking lot at Kean University
in New Jersey.
•Agent later identified as pneumonic plague.
30
TOPOFF 3
•Local hospitals involved – patients streaming
in.
•All NJ counties became Points of
Dispensing (PODS) for antibiotics.
•One POD was at the Rutgers Athletic
Center.
31
TOPOFF 3: General Observations
•Totally scripted or playbook exercise.
•Lacked random introduction of surprise or
contradictory information.
–Would ADT-generated models have helped
the designers here?
•No flexibility for game controller to change
agenda – even after the identity of the
biological agent was disclosed a week before
the event started.
32
TOPOFF 3: General Observations
•Very quick identification of the agent as plague –
less than 24 hours.
•No attempt to use array of databases to help in
identification of the agent. In smart cities, this
would be done.
–Note: Pneumonic plague takes 2-3 days before
symptoms appear
•No “chaos” of responding to
an unknown biological agent.
Pneumonic plague
in India
33
TOPOFF 3: General Observations
•Lack of truly significant random
perturbations
–Underscores importance of randomness in
modeling responses to health events; ADT
would allow much more sophisticated testing
•No inconsistent information that might lead
to refutation of initial hypothesis about cause.
–Would ADT-generated modeling have helped
develop a better exercise in this sense?
34
TOPOFF 3: General Observations
•People were being shipped off to hospitals
without any idea (in the “script”) of what the
contaminant might have been.
–Models might help us understand the
danger of such a decision.
–In real emergency, algorithms would absorb
data and help us determine where to send
people.
–Algorithms would help us
consider alternatives
Idea of quarantine on Kean
University campus was not considered.
35
TOPOFF 3: Concept of POD
•In a POD: We bring together large numbers
of people to receive their materials in one
location.
–Hand out antibiotics
–Hand out educational materials about the
disease and the medicine
•How do you get them there?
–Smart Cities: traffic congestion, parking,
etc.; models modified
in real time
–Smart cities: Instructions to people
36
TOPOFF 3: Concept of POD
•Other ADT Issues in modeling the POD:
–How do you get enough volunteers?
–How do you get food to the volunteers? The
patients?
–Who gets priority? Triage.
37
TOPOFF 3: Concept of POD
•Still other issues in modeling the POD:
–How do you handle panic within the POD?
–Pushing, shoving.
–People on long lines.
–People on lines getting sick.
–In our observation: TOPOFF 3
had none of these elements.
–Modeling challenge: social
responses to health events
–Better and more rapid
information can help avoid
panics
38
TOPOFF 3: Concept of POD
•POD Loading Issues:
–What is maximum capacity of a POD?
–How many workers are needed?
–How much time is it reasonable to keep patients there?
–How to handle short preparation time before masses of
people arrive?
–What is adequate time to screen individuals?
–How do you prevent a secondary attack if a mass of
people are gathered in one place?
–These are all modeling issues.
–Real-time data feedback could really help smart city
managers handle these kinds of questions.
39
TOPOFF 3: Concept of POD
•Some conclusions about PODS:
–The most successful POD violated the rules.
–It was a Point of Distribution, not a Point of
Dispensing.
–Medicines were distributed to a few people in
large quantities.
–They in turn redistributed the drugs to others –
away from the POD.
–Smart Cities: Massive databases; record
keeping in advance helped distributors know
where to go and to whom to give drugs
40
TOPOFF 3: Concept of POD
•Some conclusions about PODS:
–The most successful POD serviced 67,000
people in 4 hours. This was the one that wasn’t
really a POD.
–The others serviced 500 to 1000.
–Conclusion: Decentralization could be a key
avoid mass movement of people
–Advantages of dispensing drugs and
information in local communities.
–But: is decentralization always best?
–Modeling challenges
–Smart City challenge: Information
challenges under decentralization
41
TOPOFF 3: Closing Comment
•Officials in NJ and at Federal Emergency
Management Agency (FEMA) were very
interested in our observations.
•They seemed quite open to more technical
analysis of the exercise and more technical
approaches to future planning.
•Published in J. of Emergency Management
42
Example 3: ADT and Smart Grid
Many of the following ideas are borrowed from a
presentation by Gil Bindewald of the Dept. of Energy to
the SIAM Science Policy Committee, 10-28-09
43
ADT and Smart Grid
•Today’s electric power systems have grown up
incrementally and haphazardly – they were not
designed from scratch
•They form complex systems that are in constant
change:
−Loads change
−Breakers go out
−There are unexpected disturbances
−They are at the mercy of uncontrollable
influences such as weather
44
ADT and Smart Grid
•Today’s electric power systems operate under
considerable uncertainty
•Cascading failures can have dramatic
consequences.
45
ADT and Smart Grid
•The management of a “smart city” faces many
challenges in understanding and controlling the
electric power available to its citizenry and
helping to avoid catastrophic outages and
failures.
•The smart city can aid citizens in managing their
power usage through guidance and directives
based on incredibly detailed understanding of
their usage:
–Benefits individuals
–Benefits the entire city or metropolitan area
46
ADT and Smart Grid
•Power grid challenges include:
−Huge number of customers, uncontrolled
demand
−Changing supply mix system not designed for
complexity of the grid
−Operating close to the edge and thus
vulnerable to failures
47
ADT and Smart Grid
•Power grid challenges include:
−Interdependencies of electrical systems create
vulnerabilities
−Managed through large parallel computers/
supercomputers with the system not set up for
this type of management
48
ADT and Smart Grid
•Power grid advantages: “Smart grid” data
sources enable real-time precision in operations
and control previously unobtainable:
−Real-time data from smart meter systems will
enable customer engagement through demand
response, efficiency, etc.
Help understand power use
Help conserve
Help power companies
Control use
−This is a good example of a
49
service provided by a smart city
ADT and Smart Grid
•“Smart grid” data sources enable
real-time precision in operations and
control previously unobtainable:
−Time-synchronous phasor data,
linked with advanced computation
and visualization, will enable
advances in
state estimation
real-time contingency analysis
real-time monitoring of
dynamic (oscillatory) behaviors in
the system
50
ADT and Smart Grid
•“Smart grid” data sources enable real-time
precision in operations and control previously
unobtainable:
−Enhanced operational intelligence
−Integrating communications, connecting
components for real-time information and control
51
ADT and Smart Grid
•“Smart grid” data sources enable real-time
precision in operations and control previously
unobtainable:
−Sensing and measurement technologies will support
faster and more accurate response, e.g., remote
monitoring
−Advanced control methods will enable rapid
diagnosis and precise solutions appropriate to an
“event”
52
ADT and Smart Grid: Phasor
Measurements
•Phasor measurements will provide “MRI
quality” visibility of the power system.
•Traditional SCADA measurement provides
−Bus voltages
−Line, generator, and transformer flows
−Breaker Status
−Measurement every 2 to 4 seconds
53
ADT and Smart Grid: Phasor
Measurements
•Phasor technology and phasor measurements
provide additional data:
−Voltage and current phase angles
−Frequency rate of change
−Measurements taken many times a second
−This gives dynamic visibility into power system
behavior
54
ADT and Smart Grid: Phasor
Measurements
•Phasor technology and phasor measurements
provide additional data:
−New algorithmic methods to understand,
process, visualize data and find anomalies
rapidly are required.
−Such measurements will allow rapid
understanding of how customers are
using electricity: Smart meters.
−Raise privacy issues.
What movie am I watching?
−Another area for research.
55
ADT and Smart Grid
Other problems/opportunities requiring ADT:
•Grid robustness: How will the grid respond to
disturbances and how quickly can it be restored to its
healthy state?
•Transmission reliability:
−Wide area situational awareness and advanced
computational tools can help with quick
response to dynamic process changes, e.g.,
using automatic switching.
−Sample challenge: How far are we from the
edge? When voltages drop too fast, the entire
power system can collapse.
56
ADT and Smart Grid
Some Challenges:
•Need to develop reliable, robust models to help us
achieve system understanding.
•Need a new mathematics for characterizing
uncertainty in information created from the large
volumes of data arising from the smart grid.
•Need new methods to enable the use of highbandwidth networks by dynamically identifying
only the data relevant to the current information
need and discarding the rest. Similar challenges
for many smart city applications.
57
ADT and Smart Grid
Some Challenges:
•Security of new software is a priority – same for
many smart city applications
•Cyber attacks on the electric power grid are a
major concern. Needed are methods for
−Prevention
−Response
−Recovery
58
ADT and Smart Grid: Summary
Relevance of ADT
Algorithmic methods needed to aid smart cities:
•Improve security of energy system in light of its
haphazard construction and dynamically changing
character
•Find early warning of a changed state – anomaly
detection
59
ADT and Smart Grid: Summary
Relevance of ADT
Algorithmic methods needed to aid smart cities:
•Identify and overcome vulnerabilities in the
system
•Protect the privacy of individuals under new data
collection methods
60
ADT and Smart Grid: Summary
Relevance of ADT
Algorithmic methods needed to aid smart cities:
•Protect systems operating “close to the edge”
•Find new ways to characterize uncertainty in
information about the health of the system
•Find ways to protect against cyber attacks that
take advantage of vulnerabilities created by
dependence on massive amounts of data generated
through the smart grid.
•So: implementation & development of smart city
methods requires not only new research on ADT,
but research to protect against new vulnerabilities
61
our smart cities create.