copyright 2000-2003 alidade consulting all rights reserved administration restrooms business center...
Post on 18-Dec-2015
223 Views
Preview:
TRANSCRIPT
S3
T2
T1
I3
S1
S2
D2
D1
I1
2I
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
ADMINISTRATIONADMINISTRATION
• Restrooms
• Business Center
• Food and Drink
• Problems? See Betty
• ROE
• Product? Posted to www.alidade.net
• Restrooms
• Business Center
• Food and Drink
• Problems? See Betty
• ROE
• Product? Posted to www.alidade.net
New Horizons in Search Theory3nd International Workshop
NETWORKED SEARCHES AND SEARCHES IN NETWORKS
New Horizons in Search Theory3nd International Workshop
NETWORKED SEARCHES AND SEARCHES IN NETWORKS
Introduction and Recap
Jeff Cares3 September 2003
Introduction and Recap
Jeff Cares3 September 2003
S3
T2
T1
I3
S1
S2
D2
D1
I1
2I
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
New Horizons in Search Theory:From Abstraction to Application
May 21-23, 2001
Newport, Rhode Island
S3
T2
T1
I3
S1
S2
D2
D1
I1
2I
ALIDADECONSULTING
NEWPORT CENTER FOR
INFORMATION AGE WARFARE STUDIES
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
AbstractAbstract• Search Theory
– Developed when operators and academics collaborated to apply mathematics to a very tangible task (ASW)
– Early successes contributed to the foundations of OR – This field now consists of two types of professionals,
- analysts who work directly with operators- academics who develop new mathematical results.
• These two groups have diverged – OR Analysts continue to ply their trade, mostly in ASW.– Academics have gravitated to extraordinarily abstract
results in complex multivariate mathematics.
• Search Theory – Developed when operators and academics collaborated
to apply mathematics to a very tangible task (ASW)– Early successes contributed to the foundations of OR – This field now consists of two types of professionals,
- analysts who work directly with operators- academics who develop new mathematical results.
• These two groups have diverged – OR Analysts continue to ply their trade, mostly in ASW.– Academics have gravitated to extraordinarily abstract
results in complex multivariate mathematics.
Researchers working within what has come to be knownas “complexity science” have developed results that have enriched
the related fields of search theory and heuristic optimization.
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
AgendaAgenda
• RADM Sprigg, Commander, NWDC• Three Speakers with Discussion
– Introduction, Jeff Cares, Alidade Consulting– Dr. Brian McCue, Center for Naval Analyses
• “A Conversation about Searches”
– Dr. Jose Lobo, Cornell University• “New Topics in Heuristic Optimization”
– Dr. Bill Macready, Bios Group, Inc.• “The No Free Lunch Theorem”
• Dr. Martin Shubik, Yale University– “The Art, Science and Politics of Modeling
Combat”
• RADM Sprigg, Commander, NWDC• Three Speakers with Discussion
– Introduction, Jeff Cares, Alidade Consulting– Dr. Brian McCue, Center for Naval Analyses
• “A Conversation about Searches”
– Dr. Jose Lobo, Cornell University• “New Topics in Heuristic Optimization”
– Dr. Bill Macready, Bios Group, Inc.• “The No Free Lunch Theorem”
• Dr. Martin Shubik, Yale University– “The Art, Science and Politics of Modeling
Combat”
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
• Three Speakers with Discussion– Mr. Don Delbalzo, Naval Research Lab
• “GRASP”
– CDR John Dickmann, NCIAWS• “Information Age Warfare”
– Mr. Jeff Cares, Alidade Consulting• “An Information Age Combat Model”
• Stu Kauffman, M.D., Bios Group, Inc.• Colloquium
– Facilitation: Dr. Jose Lobo, Cornell University
• Prof. Wayne Hughes, NPS• “On Koopman”
• Three Speakers with Discussion– Mr. Don Delbalzo, Naval Research Lab
• “GRASP”
– CDR John Dickmann, NCIAWS• “Information Age Warfare”
– Mr. Jeff Cares, Alidade Consulting• “An Information Age Combat Model”
• Stu Kauffman, M.D., Bios Group, Inc.• Colloquium
– Facilitation: Dr. Jose Lobo, Cornell University
• Prof. Wayne Hughes, NPS• “On Koopman”
AgendaAgenda
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
A Conversation about Searches Dr. Brian McCue, Center for Naval Analyses
A Conversation about Searches Dr. Brian McCue, Center for Naval Analyses
• Overview of search theory in ASW operations– discussing physical and non-physical (i.e., ESM) searches and means
of estimating the numbers of targets that were not detected.
• For non-physical searches, mathematical techniques allow an estimate of how many targets were not detected (messages not intercepted).
• Impact of target counter-detection actions. – Led to a discussion of how the Air Force tends to treat targets as
passive objects, and of the applicability of traditional search theory to future problems with many variables and targets/sensors numbering in the thousands.
– Complexity effects such as information flow, multiple interactions, and “edge” effects (different results near physical boundaries such as land, shallow water, etc.)
• Importance of displaying negative information (areas not searched, or searched but with no detections). – Current C2 and display systems do not deliver this information to
warfighters.
• Overview of search theory in ASW operations– discussing physical and non-physical (i.e., ESM) searches and means
of estimating the numbers of targets that were not detected.
• For non-physical searches, mathematical techniques allow an estimate of how many targets were not detected (messages not intercepted).
• Impact of target counter-detection actions. – Led to a discussion of how the Air Force tends to treat targets as
passive objects, and of the applicability of traditional search theory to future problems with many variables and targets/sensors numbering in the thousands.
– Complexity effects such as information flow, multiple interactions, and “edge” effects (different results near physical boundaries such as land, shallow water, etc.)
• Importance of displaying negative information (areas not searched, or searched but with no detections). – Current C2 and display systems do not deliver this information to
warfighters.
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
Hard Search Problems: Recent Developments in Heuristic Optimization
Dr. Jose Lobo, Cornell University
Hard Search Problems: Recent Developments in Heuristic Optimization
Dr. Jose Lobo, Cornell University
• Focused on business applications of search processes. – innovation is a search
• Hard search problems where the nature of the solution space precludes the use of traditional analytical methods. – numerous local optimum – multiple conflicting constraints – problems where the payoff in unknown
• Promising developments in heuristic optimization– Intelligent algorithms – Distributed search – Agent directed adaptive searches,– Ability to learn more about the search space.
• Focused on business applications of search processes. – innovation is a search
• Hard search problems where the nature of the solution space precludes the use of traditional analytical methods. – numerous local optimum – multiple conflicting constraints – problems where the payoff in unknown
• Promising developments in heuristic optimization– Intelligent algorithms – Distributed search – Agent directed adaptive searches,– Ability to learn more about the search space.
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
Multi-Paradigm OptimizationWilliam Macready, Bios Group
Multi-Paradigm OptimizationWilliam Macready, Bios Group
• Matching the biases inherent in an algorithm with the nature of the problem being addressed. – Assumptions built into an algorithm are often implicit and unknown
to the user.
• Problem solving process frequently starts with an algorithm, which is matched to a problem. – Process needs to start with the problem, for which an appropriate
algorithm can then be selected.
• Combining algorithms could be applied to:– Very complex objective functions (complex simulations)– Where multi-level models are available– When model inputs and outputs may not be exactly known– When many good solutions are necessary or problems change
• Matching the biases inherent in an algorithm with the nature of the problem being addressed. – Assumptions built into an algorithm are often implicit and unknown
to the user.
• Problem solving process frequently starts with an algorithm, which is matched to a problem. – Process needs to start with the problem, for which an appropriate
algorithm can then be selected.
• Combining algorithms could be applied to:– Very complex objective functions (complex simulations)– Where multi-level models are available– When model inputs and outputs may not be exactly known– When many good solutions are necessary or problems change
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
The Art, Science and Politics of Modeling Combat
Dr. Martin Shubic, Yale University
The Art, Science and Politics of Modeling Combat
Dr. Martin Shubic, Yale University
• The importance of art, science and politics in getting things done in defense circles. – cost uncertainties (including accounting fudge factors)– risk aversion – networking and of setting achievable goals
• Growth of technology has created an “opaqueness in understanding what we are doing.” – equations that no one has seen or can understand. – really complex simulation systems can be manipulated to
produce a good answer to justify anything.
• The art of modeling includes:– the selection of the key variables and relationships– importance of operational experience
• The importance of art, science and politics in getting things done in defense circles. – cost uncertainties (including accounting fudge factors)– risk aversion – networking and of setting achievable goals
• Growth of technology has created an “opaqueness in understanding what we are doing.” – equations that no one has seen or can understand. – really complex simulation systems can be manipulated to
produce a good answer to justify anything.
• The art of modeling includes:– the selection of the key variables and relationships– importance of operational experience
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
Genetic Range Dependent Algorithm for Search Planning (GRASP)
Don Debalzo, NRL
Genetic Range Dependent Algorithm for Search Planning (GRASP)
Don Debalzo, NRL
• ASW search planning TDA – uses current environmental data to calculate
sensor performance and optimum search tracks for up to several heterogeneous sensors.
– designed to aid in sensor selection, placement, and utilization at the platform and battle group level.
– basic target behaviors can be incorporated – used a series of at sea ASW exercises.
• The GRASP algorithm randomly adjusts waypoints in candidate paths to produce the optimum search track over time.
• ASW search planning TDA – uses current environmental data to calculate
sensor performance and optimum search tracks for up to several heterogeneous sensors.
– designed to aid in sensor selection, placement, and utilization at the platform and battle group level.
– basic target behaviors can be incorporated – used a series of at sea ASW exercises.
• The GRASP algorithm randomly adjusts waypoints in candidate paths to produce the optimum search track over time.
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
ColloquiumColloquium
• Is Search Still Relevant? – One participant suggested that if we succeed in blanketing the
battlespace with a very large numbers of sensors, we might be able to apply the asset allocation model for search planning
– examine new search theory developments in light of value added -- i.e., what do the new techniques allow us to do that we could not do before? What new insights are added?
• How is Search Initiated? – The thrust of this question was whom in future combat
organizations will determine that unfulfilled information requirements exist and direct a search effort.
– application of agent-based models to determine how an enemy is operating based on observations.
– How to validate theories on enemy force dispositions, organization and operating patterns, and that such a system could automate the process of intelligence fusion.
• Is Search Still Relevant? – One participant suggested that if we succeed in blanketing the
battlespace with a very large numbers of sensors, we might be able to apply the asset allocation model for search planning
– examine new search theory developments in light of value added -- i.e., what do the new techniques allow us to do that we could not do before? What new insights are added?
• How is Search Initiated? – The thrust of this question was whom in future combat
organizations will determine that unfulfilled information requirements exist and direct a search effort.
– application of agent-based models to determine how an enemy is operating based on observations.
– How to validate theories on enemy force dispositions, organization and operating patterns, and that such a system could automate the process of intelligence fusion.
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
• What does it take to bring scientific rigor the heuristic algorithms? – Dr. Lobo replied that heuristics algorithms should be treated
as an experimental science. – He noted that by definition, HAs are used in case where
traditional analytic methods do not work. – He added that continued improvements were needed in
documenting and reporting research on HAs in a manner consistent with other experiments.
– He also said that more research is needed to explore why some algorithms work well on some problems but not on others.
• Discussion of inability of most military personnel to understand how algorithms, and TDAs based on algorithms actually work.
• What does it take to bring scientific rigor the heuristic algorithms? – Dr. Lobo replied that heuristics algorithms should be treated
as an experimental science. – He noted that by definition, HAs are used in case where
traditional analytic methods do not work. – He added that continued improvements were needed in
documenting and reporting research on HAs in a manner consistent with other experiments.
– He also said that more research is needed to explore why some algorithms work well on some problems but not on others.
• Discussion of inability of most military personnel to understand how algorithms, and TDAs based on algorithms actually work.
ColloquiumColloquium
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
• The group noted three broad issues surfaced repeated, and that these may be good themes around which to organize. The issues were:– Network design– Use of heuristic algorithms to solve real world
problems– Application of computer science tools to combat
simulations
• Applications beyond the Navy were also discussed. – USCG search and rescue, Army mine clearance,
and Air Force TAMD attack operations
• The group noted three broad issues surfaced repeated, and that these may be good themes around which to organize. The issues were:– Network design– Use of heuristic algorithms to solve real world
problems– Application of computer science tools to combat
simulations
• Applications beyond the Navy were also discussed. – USCG search and rescue, Army mine clearance,
and Air Force TAMD attack operations
ColloquiumColloquium
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
S3
T2
T1
I3
S1
S2
D2
D1
I1
2I
New Horizons in Search Theory 2nd International Workshop
“Search Theory, Invented Now”
NEWPORT CENTER FOR
INFORMATION AGE WARFARE STUDIES
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
• Third claimant to search solutions is emerging– Almost every advanced warfare concept depends on high-
performance collective robotics, most notably for search and surveillance operations.
• This raises an intriguing proposition– Might arranging the instruments of a search (collective
autonomous agents) require the same complex optimization solutions as the process of the search (finding the target)?
• If so, we are on the verge of a new, ironic definition of Search Theory– More importantly, it appears that the fundamental precepts
of Koopman-style Search Theory (KST) might warrant reconsideration
• Third claimant to search solutions is emerging– Almost every advanced warfare concept depends on high-
performance collective robotics, most notably for search and surveillance operations.
• This raises an intriguing proposition– Might arranging the instruments of a search (collective
autonomous agents) require the same complex optimization solutions as the process of the search (finding the target)?
• If so, we are on the verge of a new, ironic definition of Search Theory– More importantly, it appears that the fundamental precepts
of Koopman-style Search Theory (KST) might warrant reconsideration
IntroductionIntroduction
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
Search Theory, Invented NowSearch Theory, Invented Now
• Might be similar to KST– Might use Monte Carlo techniques, perhaps involve more statistics, or it
might use functional forms that require numerical solution by modern computers.
– It might think of the search as updating a Bayesian "prior" as to where the target isn't. In other words, it might look a lot like what people are
doing today to modify KST.
• Might look dramatically different in other ways– Searchers and agents, applying a selection of local rules, sometimes
changing mode or state so as to switch to a new menu of rules?
– Might the robotic elements of a future force act as such agents?
– Koopman’s later editions of Search and Screening included an exploration of Information Theory, a topic unknown before World War II. How might Information Theory contribute to a modern Search Theory?
• Might be similar to KST– Might use Monte Carlo techniques, perhaps involve more statistics, or it
might use functional forms that require numerical solution by modern computers.
– It might think of the search as updating a Bayesian "prior" as to where the target isn't. In other words, it might look a lot like what people are
doing today to modify KST.
• Might look dramatically different in other ways– Searchers and agents, applying a selection of local rules, sometimes
changing mode or state so as to switch to a new menu of rules?
– Might the robotic elements of a future force act as such agents?
– Koopman’s later editions of Search and Screening included an exploration of Information Theory, a topic unknown before World War II. How might Information Theory contribute to a modern Search Theory?
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
SessionsSessions
• Jeff Cares, Alidade– Recap of last year– Statements of interest
• Dr. Brian McCue, CNA– Midway Game– Very “agent-looking”– Dynamic co-evolution– Intent very difficult to discern from knowledge
of physical elements
• Jeff Cares, Alidade– Recap of last year– Statements of interest
• Dr. Brian McCue, CNA– Midway Game– Very “agent-looking”– Dynamic co-evolution– Intent very difficult to discern from knowledge
of physical elements
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
SessionsSessions
• Dr. Brian McCue, CNA– “Search Theory, Invented Now”– Recognition of KST as a product of its time– Suggestion that problem may be more complex
• CDR Mike Rimmer, RN, NWDC– “Future Concepts of Military Search”– Expeditionary Sensor Grid, FORCEnet, FCS, etc.– Formulation of the operators’ estimates of the
problem
• Dr. Brian McCue, CNA– “Search Theory, Invented Now”– Recognition of KST as a product of its time– Suggestion that problem may be more complex
• CDR Mike Rimmer, RN, NWDC– “Future Concepts of Military Search”– Expeditionary Sensor Grid, FORCEnet, FCS, etc.– Formulation of the operators’ estimates of the
problem
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
SessionsSessions
• Dr. Doug Gage, DARPA-SDR– “Distributed Robotics and Problems of Search”– Review of existing DARPA efforts with robotic
collective– Discussion of engineering problems
• Hard for engineers to build ‘em cheap
• Dr. Fred Glover, OptTek, Inc., U of Colo.– “TABU Search”– Discussion of TABU search methods and
effectiveness
• Dr. Doug Gage, DARPA-SDR– “Distributed Robotics and Problems of Search”– Review of existing DARPA efforts with robotic
collective– Discussion of engineering problems
• Hard for engineers to build ‘em cheap
• Dr. Fred Glover, OptTek, Inc., U of Colo.– “TABU Search”– Discussion of TABU search methods and
effectiveness
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
SessionsSessions
• Dr. Ralph Klingbeil, NUWC– “Search with Distributed Forces”– Summary of research into dimensions of the problem– Rudimentary exploration of the contributions of
distributed forces
• Cpt Alistair Dickie, RAA– “Multi-Agent Robot Swarm Simulation”– Discussion of his research at NPS– Demonstration of his simulation
• Dr. Ralph Klingbeil, NUWC– “Search with Distributed Forces”– Summary of research into dimensions of the problem– Rudimentary exploration of the contributions of
distributed forces
• Cpt Alistair Dickie, RAA– “Multi-Agent Robot Swarm Simulation”– Discussion of his research at NPS– Demonstration of his simulation
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
SessionsSessions
• LtCol Ray Hill, AFIT– “Agent-Based Analysis of U-Boats in Bay of Biscay”– ABM treatment of (new) classic McCue work– Raised interesting questions about validation of ABM
results
• Dr Josh Epstein, Brookings Institute– “The Spread of Disease in Populations”– Recent work for NSF– Explored the dynamics of contagion, death and
recovery
• LtCol Ray Hill, AFIT– “Agent-Based Analysis of U-Boats in Bay of Biscay”– ABM treatment of (new) classic McCue work– Raised interesting questions about validation of ABM
results
• Dr Josh Epstein, Brookings Institute– “The Spread of Disease in Populations”– Recent work for NSF– Explored the dynamics of contagion, death and
recovery
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
SessionsSessions
• Dr. Rob Axtell, Brookings Institute– “Operations Research as a Subset of MAS”– Intentionally provocative suggestion– Explored the usefulness of MAS simulations when
closed form equations fail
• Dr. Alfred Brandstein, MCCDC/Mary McDonald, SAIC– “Search Theory in Light of the El Farol Problem”– “Co-evolving Strategies between Red and Blue”– Discussion of the competitive dynamics of Search
• Dr. Rob Axtell, Brookings Institute– “Operations Research as a Subset of MAS”– Intentionally provocative suggestion– Explored the usefulness of MAS simulations when
closed form equations fail
• Dr. Alfred Brandstein, MCCDC/Mary McDonald, SAIC– “Search Theory in Light of the El Farol Problem”– “Co-evolving Strategies between Red and Blue”– Discussion of the competitive dynamics of Search
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
ColloquiumColloquium
• VV&A– How can we VV&A Agent-Based Models?– If we can’t VV&A with Blue and Red dots, how will we
gain confidence in new C2 systems for distributed search?
• Engineering of Distributed Systems– As the Search Theory member of an engineering
team building robotic systems, what recommendations should be made to the team?
– Consensus was that our knowledge of the problem is too immature to make such recommendations
• VV&A– How can we VV&A Agent-Based Models?– If we can’t VV&A with Blue and Red dots, how will we
gain confidence in new C2 systems for distributed search?
• Engineering of Distributed Systems– As the Search Theory member of an engineering
team building robotic systems, what recommendations should be made to the team?
– Consensus was that our knowledge of the problem is too immature to make such recommendations
New Horizons in Search Theory, 3nd Workshop
“NETWORKED SEARCHES AND SEARCHES IN NETWORKS”
ALIDADEINCORPORATED
NEWPORT CENTER FOR
INFORMATION AGE WARFARE STUDIES
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
Two New ProblemsTwo New Problems
• Networked Search– High-performance Collective Robotics– Swarms, Sensor Grids, etc.
• Search in Networks– Data Mining– Distributed Problem Solving
• Including Search
• Networked Search– High-performance Collective Robotics– Swarms, Sensor Grids, etc.
• Search in Networks– Data Mining– Distributed Problem Solving
• Including Search
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
AgendaWednesday
AgendaWednesday
• Morning– Intro/Recap: Jeff Cares– Distributed Networked Forces: Ray Christian
• Lunch– Dave Davis
• Afternoon– Another View of the Small World: Brian
McCue– Agent Searches in the Bay of Biscay: Ray Hill
• Dinner– Peter Dodd
• Morning– Intro/Recap: Jeff Cares– Distributed Networked Forces: Ray Christian
• Lunch– Dave Davis
• Afternoon– Another View of the Small World: Brian
McCue– Agent Searches in the Bay of Biscay: Ray Hill
• Dinner– Peter Dodd
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
AgendaThursday
AgendaThursday
• Morning– Colloquium
• Lunch– Working
• Afternoon– Colloquium
• Dinner– Hosted, 31 Willow Street
• Morning– Colloquium
• Lunch– Working
• Afternoon– Colloquium
• Dinner– Hosted, 31 Willow Street
Copyright 2000-2003 Alidade Consulting
All Rights Reserved
AgendaFriday
AgendaFriday
• Morning– Workshop, Ayrault House
• Lunch– Working
• Afternoon– Workshop, Ayrault House
• Morning– Workshop, Ayrault House
• Lunch– Working
• Afternoon– Workshop, Ayrault House
top related