Michael R. Bartolacci, Albena Mihovska, and Dilek Ozceylan on "Optimization Modeling and Decision Support for Wireless Infrastructure Deployment in Disaster Planning and Management" at ISCRAM 2013 in Baden-Baden.
10th International Conference on Information Systems for Crisis Response and Management
12-15 May 2013, Baden-Baden, Germany
Handwritten Text Recognition for manuscripts and early printed texts
Optimization Modeling and Decision Support for Wireless Infrastructure Deployment in Disaster Planning and Management
1. Optimization Modeling and
Decision Support for
Wireless Infrastructure
Deployment in Disaster
Planning and Management
Michael R. Bartolacci
Penn State University – Berks, U.S.
Albena Mihovska
Aalborg University, Denmark
Dilek Ozceylan
Sakarya University, Turkey
2.
3. Natural and Manmade Disasters
Create Problems for Wireless
Networks
Damaged wireless mobile network base
stations (towers and associated equipment)
(e.g. – 29,000 base stations were affected by
the 2011 Tohoku Earthquake and Tsunami)
Damaged mobile switching centers
Damaged “landline” connectivity (coaxial and
fiber optic cable networks) that interface with
the wireless networks (e.g. - 1.9 million
fixed-line service subscribers were affected
by the 2011 Tohoku Earthquake and
Tsunami)
5. Impact from 2011 Tohoku
Earthquake and Tsunami
Red areas are NTT DoCoMo service areas disrupted by the disaster
and gray areas are fixed-line NTT East service areas disrupted by
the disaster in three Japanese prefectures (Iwate, Myagi, and
Fukushima with the earthquake location shown)
7. Wireless Infrastructure Planning
and Management During Disasters
The Impacted Organizations
◦ Local, state, federal governmental agencies that
own/lease portable wireless network devices and
mobile infrastructures and provide disaster relief
◦ Mobile network providers with existing subscribers
or connectivity in an affected region
◦ Fixed line telecommunications providers in an
affected region
◦ NGO’s that utilize wireless networks to provide
disaster relief
◦ Public utility companies and general commerce
8. Wireless Infrastructure and its
Facilitation of Social Media are
Playing Increasing Roles in
Emergency Management
For identifying/locating family, friends, etc.
For coordinating emergency response
For identifying/locating “persons of interest”
(such as in Boston recently)
For adding to the resilience of an affected
populace by allowing people to help
themselves and their neighbors in real time
(such as informing them where emergency
supplies will be distributed)
9. Dominant Uses of Social Media During
the 2011 Tohoku Earthquake and
Tsunami (as per a World Bank Report)
Social media requires the network infrastructure to be
operable
10. In the Context of Last Year’s
ISCRAM Presentation and Beyond
Ozceylan and Bartolacci looked at the impact of the availability of
wireless mobile connectivity on the resilience of a populace
affected by a disaster (specifically looking at China and other
developing countries)
World Bank Report on telecommunications related to the 2011
Tohoku Earthquake recommended improving the reliability of
communication networks in developing countries in the context of
natural disasters by:
◦ Reducing damage by developing backup systems, such as
batteries, generators, and backup trunk lines
◦ Mitigating congestion by increasing the capacity of facilities such as
switching equipment
◦ Restoring service by deploying emergency facilities, such as portable
switching equipment and portable satellite stations
◦ All three have implications for modeling both the planning and
management of a portable wireless infrastructure
11. The Reality of Replacing Wireless
Infrastructure
Talk given at Wireless Telecommunications
Symposium 2011 (WTS 2011) in New York City in
April 2011 by an official from NTT DoCoMo (Japan’s
leading wireless mobile network operator)regarding
the aftermath of the 2011 Tohoku Earthquake and
Tsunami in March 11, 2011
◦ The general tone of the talk spoke to the need to react
quickly and repair or replace several thousand non-
functioning or destroyed base stations in the impacted
areas, some with portable ones
◦ He did not put forth any preplanning for such a deployment
and the assignment of mobile base stations in the affected
areas appeared to be conducted in an ad hoc fashion
12. The Reality of Replacing Wireless
Infrastructure
Currently NTT DoCoMo has only about 50 truck-
operated portable base stations with 3G capabilities
and is currently expanding its inventory of 4G ones
The ability of a wireless network operator to come
back online quickly or be “robust” in the face of a
disaster may even lead to greater customer loyalty or
increase subscriber base
http://japandailypress.com/ntt-docomo-to-deploy-
truck-based-lte-base-stations-0126160
During Superstorm Sandy, I went without power for 5
days and without mobile network service for 3.5; I
have mobile devices with two different carriers and
AT&T stayed functional longer and came back online
before Verizon, but I am unsure as to why
13. The Reality of Replacing Wireless
Infrastructure
Softbank, Japan’s third largest mobile
network operator, lost 3,786 base stations to
the 2011 Tohoku Earthquake and Tsunami
http://www.softbank.co.jp/en/initiatives/csr/r
econstruction/instance_01/contents_01/
The are installing new base stations with
extended life batteries (but they only last for
24 hours)
16. An Innovative Way to Provide
Temporary Wireless Connectivity
Use of small helium-filled blimps for providing
temporary wireless mobile network
connectivity
Such balloon-based base stations would have
a 3 km radius of coverage
Softbank, Japan’s third largest wireless
mobile network operator is already testing
this platform for emergency use
http://japandailypress.com/softbank-develops-blimps-for-
floating-emergency-cell-towers-112047
17. An Innovative Way to Provide
Temporary Wireless Connectivity
Conversion of voice transmissions (short
messages) to text and sent over the
network as data packets (was done during
the Japanese disaster in 2011)
Use of VSAT technology (Very Small
Aperture Terminal) which uses a small
satellite dish and a LAN to connect small
mobile terminals to the Internet and for
voice communications
18. Emerging Network Standards
That Can Be Applicable
SON (Self Organizing Networks) – grew out
of the 3GPP (Third Generation Partnership
Project) and created standards for networks
that self organize and self “heal” (deal with
link and node failures)
LTE (Long Term Evolution) heterogeneous
networks (HetNet) allows for the deployment
of picocells with little planning in terms of
their integration into the network
19. Why Do Wireless Technologies
Involved Matter ?
Conference reviewer questioned why the
wireless network technologies should be
included in the modeling discussion
◦ Limited resource for any model developed is
the number of portable base stations available;
and due to the variety of technologies that may
be employed, it imposes limits on the model
◦ There are also opportunities for the inclusion of
innovative technologies (such as the blimp-
based base stations) into the modeling effort
20. Why Do Wireless Technologies
Involved Matter ?
Conference reviewer questioned why the wireless
network technologies should be included in the
modeling discussion
◦ Deployment of emerging technologies, such as hydrogen
fuel cell powered base stations that could still operate
when their power grid sources fail, may be factored into
such a model (total cost to deploy a portable base
station and lost service until it is deployed against the
cost of the fuel cell powered one)
◦ Cost to retrofit existing base stations with generators
and seismic reinforcement versus the cost to deploy
portable ones and the loss of service
21. We Examined the Literature for
Models Addressing This Area
Two Types of Wireless Infrastructure Modeling
– Deterministic and Stochastic
◦ Deterministic assumes stable (in other
words, known) demand for capacity for each area
served in the network over time and usually
assumes fixed base station locations; the bulk of
the literature for wireless infrastructure modeling
assumes fixed base stations and known demand
◦ Stochastic allows for variations in capacity demand
and also variations in the topological design of the
infrastructure (such as the lack of any fixed
structure with an Ad Hoc Mobile network)
22. We Examined the Literature for
Models Addressing This Area
Two Phases to the deployment of wireless
equipment in a disaster context
◦ Planning
◦ Management
There is a plethora of wireless infrastructure
Planning models in the literature; though
none of them specifically address their use
during or after a disaster
23. We Examined the Literature for
Models Addressing This Area
Planning Models
◦ The only models in the literature examine the
expansion of capacity to new areas, adding
capacity to existing areas, or the assignment of
base stations to switching centers (which
connect to several base stations at once)
◦ Essentially, they all examine the Planning
phase in the context of normal day-to-day
operations
24. Planning Problem in a Disaster
Context
Root PLANNING problem is similar to the traditional Fixed
Base Station Location Problem (deterministic in nature)
› Given a set of candidate sites for mobile base station locations and the
associated costs to use and connect each base station to form the
wireless network architecture
› Given a set of demand constraints (users in areas that must be served by
the architecture)
› Choose the optimal design (locations for mobile base
stations, microcells, picocells, etc.) for the network that minimizes
overall cost while satisfying demand (which may include minimum
acceptable quality of service constraints or reflect a maximization of
levels of connectivity to serve)
› Can be solved before a disaster for areas that may be impacted
(essentially deciding where to place mobile base stations to cover a given
area as if no previous infrastructure existed)
25. Generalized Combined
Optimization Problem for Planning
and Management
Given an area affected by natural or manmade disaster:
› Provide wireless connectivity for the duration of the relief effort
in the form of portable base stations, microcells, picocells, etc.
(and possibly related ad hoc equipment) to support relief and
recovery efforts while maximizing connectivity (being able to get
a usable signal to communicate across the affected
area), minimizing the cost of deployment (or other similar goal)
› Subject to:
Budgetary costs constraints
Availability of base station constraints over time (stochastic)
Area connectivity coverage constraints over time (stochastic)
Mobile base station location restriction constraints
Portable Base station type match constraints
Demand (capacity) constraints over time (stochastic)
Fixed line connection point constraints
Ad hoc architectural component constraints (could be stochastic)
26. Planning Problem in a Disaster
Context
Additional Factors in the Context of Disaster PLANNING
that May Be Incorporated
The provisioning process for mobile base stations and related
portable wireless infrastructure by network
providers, governmental agencies, etc. may play a big role in
the practical application of any optimization model’s results and
could be included
Provisioning directly relates to the availability of the mobile
base stations and related equipment from both a spatial and a
temporal point of view; it forms the foundation for any cost
constraints on their deployment and use (the more money you
use to acquire an inventory of mobile base stations and
transport them to staging or storage areas, the less you have
available for the actual setup and use when and where they
are needed)
27. Management Problem in a Disaster
Context
Additional Factors in the Context of Disaster MANAGEMENT
That May Be Incorporated
Some remnants of a pre-existing functioning cellular
network may exist to incorporate into a overall design for
the deployment of mobile base stations (for a network
operator attempting to restore service), would require
incorporation into the planning model while managing its
implementation
There may be little functioning fixed line infrastructure
remaining to connect to for data backhaul or PSTN
(Public Switched Telephone Network) connectivity(for
rural and “hard hit” areas)
28. Management Problem in a Disaster
Context
Additional Factors in the Context of Disaster MANAGEMENT
That May Be Incorporated
Due to terrain, land ownership, and similar
factors, during implementation, desired connectivity
coverage may require the integration of an ad hoc
network architecture with the mobile base station cellular
architecture, creating what is termed a “multi-hop
cellular network” – not addressed in typical base station
optimization models in the literature
Only seen in a few traffic engineering models dealing
with protocol design for interoperability - “Media
Handling for Multimedia Conferencing in Multihop Cellular
Networks” – Khedher, Glitho, and Dssouli (2009)
29. Approaches In The Literature
Fixed Base Station Location PLANNING Problem with
Assignment of BS’s to Mobile Switching Centers/Controllers
included either as decision variables or incorporated into
location constraints – most of these use a two step
optimization approach where towers are first placed in an
optimal fashion and then assigned to switching centers
◦ “Location Area Planning in Cellular Networks Using Simulated
Annealing” – Demirkol, Ersoy, Caglayan, and Delic (2001)
◦ “Location Area Planning and Cell-to-Switch Assignment in Cellular
Networks” – Demirkol, Caglayan, and Delic (2004)
◦ “UMTS base station location planning: a mathematical model and
heuristic optimisation algorithms” – Yang, Aydin, Zhang, and Maple
(2007)
modeled as a p-median problem and used three meta-heuristics for
solving: genetic algorithm, simulated annealing and evolutionary-
simulated annealing
30. Approaches In The Literature
“Multiperiod Cellular Network Design via Price-
Influenced Simulated Annealing (PISA)” – Menon
and Amiri (2006)
Takes a broader temporal view of the PLANNING
problem and utilizes a hybrid heuristic with ideas
from simulated annealing and linear programming
31. Approaches In The Literature
More Base Station Placement Optimization
Problems in the Literature
◦ “Radio planning and coverage optimization of 3G
cellular networks’ – Amaldi, Capone, and Malucelli
(2008)
Includes more technical aspects of the base station’s
coverage characteristics in the optimization model
◦ “Robust Tower Location for Code Division Multiple
Access Networks” – Rosenberger and Olinick (2006)
Used a stochastic integer programming approach to
optimize locations under uncertainty of demand
32. Factors to Consider that are
Missing in the Literature
Capacity/Coverage for a given area can be
limited by the network operator for a given type
of service
◦ In other words, certain kinds of voice and data traffic can be
given priority therefore reducing the capacity needed to
cover a given area (emergency traffic can be given priority
such as in the Japanese disaster where regular traffic on the
mobile networks available was reduced by 70% and fixed-
line regular traffic was reduced by 90%)
Cost/ability to implement substitution
technologies such as satellite communications
33. Factors to Consider that are
Missing in the Literature
Optimal linking of ad hoc infrastructures to
portable base stations (ala a VSAT
arrangement, but with a terrestrial connection
instead of satellite)
Optimal preplanning of portable base station
placement based on topography and other
physical factors (it is possible to locate candidate
sites ahead of time in areas prone to disasters)
34. Factors to Consider that are
Missing in the Literature
Notion of what is “connectivity”
◦ Coverage for a proposed base station location is usually
assumed based on signal strength with little thought
given to the technology involved or related factors
Time – very little exists in the literature with
respect to planning for changes in demand
(usage) over time
◦ Creates the need to deal with the mobility of portable
base stations
35. Possible Optimization Solution
Approaches
Quasi-Static
◦ Assumes that demand is relatively static in all given
regions over a short time window and optimize a model
for the time window (probably using a suboptimal
heuristic due to computing time necessities), this
approach is a good compromise in order to reduce
complexity and time to solve, but requires re-
optimization periodically as conditions change
Stochastic
◦ A dynamic programming approach gives a more cogent
approach to solving the problem, but can involve making
many assumptions about future states in order to derive
the model adequately enough
36. Next Steps
Formulate a model with both Planning and
Management variables
Choose a solution approach
Attempt to solve to optimality if
possible, otherwise heuristically
◦ Possibly a decomposition is required for solution
Test the model with actual data and
simulation if possible