© 2006 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or

for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be

obtained from the IEEE.

For more information please see www ieee org/portal/pages/about/documentation/copyright/polilink html

www.computer.org/intelligent

The “Helpful Environment”: GeographicallyDispersed Intelligent Agents That Collaborate

Austin Tate

Vol. 21, No. 3May/June 2006

This material is presented to ensure timely dissemination of scholarly and technicalwork. Copyright and all rights therein are retained by authors or by other copyrightholders. All persons copying this information are expected to adhere to the termsand constraints invoked by each author's copyright. In most cases, these works

may not be reposted without the explicit permission of the copyright holder.

MAY/JUNE 2006 1541-1672/06/$20.00 © 2006 IEEE 57Published by the IEEE Computer Society

T h e F u t u r e o f A I

The “HelpfulEnvironment”: Geographically DispersedIntelligent Agents ThatCollaborate

Austin Tate, Artificial Intelligence Applications Institute, University of Edinburgh

A I’s first 50 years have given us powerful techniques and tools, some which have

found significant and valuable application. AI technology helps many people

on a regular basis, both directly and indirectly, through the goods they use, through the ser-

vices they receive, and in the course of their work. The promise of ubiquitous computing,

sensor grids, home robots, and Web services is anexciting new driver for AI that should see its reachextend still further into our everyday lives. AI’s rolein underpinning much of the emerging SemanticWeb is one example already of how widely we’ll usethe methods in the future.

Imagine an environment where sophisticated sen-sors and autonomous or semiautonomous diagnosis,protection, and repair systems are integral to cloth-ing, communications devices, vehicles, transporta-tion systems, buildings, and the environment. Thesewould form the basis for a distributed, adaptable, andresilient “safety net” for every individual and orga-nization at personal, family, business, regional,national, and international levels.1 In natural-disas-ter-prone areas, government legislation, buildingcodes, and insurance requirements would ensure thatall future PDAs, communication devices, vehicles,and buildings include appropriate sensor and actua-tor systems to assist both their users and othersnearby. Systems would adapt and respond to emer-gencies whether or not communication were possi-ble. Where feasible, local help would be used, withappropriate calls on shared services facilitated, when-ever this is both possible and necessary. Through thisframework, requests for assistance could be validatedand brokered to available and appropriate services ina highly distributed market fashion. Services wouldbe provided to individuals or communities throughthis network to add value and give all sorts of assis-tance beyond the emergency response aspects. In

emergency situations, the local infrastructure wouldbe augmented by the facilities of the responder teamsat any level from local police, ambulance, and fireresponse, all the way up to international response. Anemergency zone’s own infrastructure could be aug-mented on demand by laying down temporary low-cost sensor grids and placing specialized devices androbotic responders into the disaster area.

Emergency response challengesThe United Nations Office for the Coordination

of Humanitarian Affairs (http://ochaonline.un.org)is one of the international bodies that are chargedwith assisting in international crises. OCHA’s pri-mary functions are to

• develop common strategies for response,• assess situations and needs,• convene coordination forums,• mobilize resources,• address common problems, and• administer coordination mechanisms.

These challenges face any group or organization thatintends to help in a crisis.

Local or regional governments are often respon-sible for the event handling, planning, coordination,and status reporting involved in emergency response.They must harness local response capabilities to aug-ment their own by calling on other resources. Figure1 shows the Tokyo Metropolitan Government’s emer-

A future network of

sophisticated sensors,

protection, and repair

systems could be

integral to clothing,

communications

devices, transportation

systems, buildings, and

the environment.

These would form the

basis for a distributed,

adaptable, and

resilient “helpful

environment.”

gency response center, which, in the event ofan emergency, will provide information andsupport to the public (through emergencyphone lines), responders, and decision-makingauthorities.

Across a range of emergency responsescenarios, there is a common set of require-ments that intelligent computer systemsmight address. Such requirements include

• sensor data management and fusion,• accurate information gathering,• correlation and validation,• relevant and understandable commu-

nication,• contact making,• requests for assistance and matching to

available capabilities,• use of standard operating procedures and

alarms,• planning and coordination, and• scalability and robustness.

AI challengesTo give such support and make the vision

a reality, we must address many AI chal-lenges. Hiroaki Kitano and Satoshi Tadokorooutlined some of these challenges in a 50-year program of work for future rescue robot-ics in Japan.2 They also introduced the annualRoboCup Rescue Simulation competition,which tests systems in a simulation of the1995 Kobe earthquake. Several other pro-posals for “grand challenges” in computingand AI take the theme of emergencyresponse.1 For example, the UK AdvancedKnowledge Technologies (AKT) program

(http://aktors.org) addresses emergencyresponse challenge problems,and a Europeanprogram called OpenKnowledge (http://openk.org) has chosen emergency response asone of its core challenge problems. The UKFireGrid project (http://firegrid.org) seeks tolink sophisticated, large-scale sensor gridsand faster-than-real-time simulations to emer-gency response coordination systems.

We can outline several core technologies,many with essential AI components, that mustdevelop, mature, and be integrated with othersystems to make this vision of a connectedworld a reality. Such technologies include

• sensors and information gathering (suchas sensor facilities and large-scale sensorgrids, human and photographic intelli-gence gathering, information and knowl-edge validation and error reduction, theSemantic Web and metaknowledge, sim-ulation and prediction, data interpretation,and identification of “need”),

• emergency response capabilities and avail-ability (such as matching needs, broker-ing, and “trading” systems; agent tech-nology for enactment, monitoring, andcontrol; and provision of robust multi-modal communications),

• hierarchical, distributed, large-scale sys-tems (such as local versus centralized deci-sion making and control, mobile and sur-vivable systems, human and automatedmixed-initiative decision making, and trustand security),

• common operating methods (such asshared information and knowledge bases;

shared standards and interlingua; sharedhuman-scale self-help Web sites and col-laboration aids; shared standard operatingprocedures at local to international levels;and standards for signs, warnings, and soforth), and

• public education (such as publicity mate-rials, self-help aids, training courses, andsimulations and exercises).

Running through all these technologies isthe need for flexible, extendible representa-tions of knowledge with rapidly alteringscope and with changing versions and refine-ments. We can’t monolithically agree on asingle representation of all the knowledgethat will be involved. The science and tech-nology of ontologies and their managementwill be vital to sustain this knowledge.

FireGridThe technologies outlined in the previous

section come from a number of fields—somemore mature than others, each with its ownphilosophy and assumptions. However, thetechnological and research advances neces-sary to realize this vision are starting to bemade. One project that’s helping to developsuch an integrated, interdisciplinary approachis FireGrid (http://firegrid.org).3 This UK pro-ject addresses emergency response in the builtenvironment, where sensor grids in large-scalebuildings are linked to faster-than-real-timegrid-based simulations and help fire respon-ders work with the building’s internal responsesystems and occupants to form a team to dealwith the emergency.

FireGrid will integrate several core tech-nologies, extending them where necessary(see figure 2). These include

• high-performance computing involvingfire models and structural models;

• wireless sensors in extreme conditionswith adaptive routing algorithms, includ-ing input validation and filtering;

• grid computing, including sensor-guidedcomputations, mining data streams for keyevents, and reactive priority-based sched-uling; and

• command and control using knowledge-based planning techniques with user guid-ance.

I-RescueThis project (http://i-rescue.org) explores

AI planning and collaboration methods inrapidly developing emergency response and

T h e F u t u r e o f A I

58 www.computer.org/intelligent IEEE INTELLIGENT SYSTEMS

Figure 1. A typical emergency response coordination center (at the TokyoMetropolitan Government).

MAY/JUNE 2006 www.computer.org/intelligent 59

rescue situations. The overall aim is the cre-ation and use of task-centric virtual organiza-tions involving people, government and non-governmental organizations, automatedsystems, the Grid, and Web services workingalongside intelligent robotic, vehicle, build-ing, and environmental systems to respond tovery dynamic events on local to global scales.

The I-X system and I-Plan planner providea framework for representing, reasoning about,and using plans and processes in collaborativecontexts.4 An underlying ontology, <I-N-C-A>, is the basis of a flexible and extendiblerepresentation for the issues and questions toaddress, nodes and activities to perform, con-straints to maintain, and annotations to keep.5

The I-X approach to plan representation anduse relates activities to their underlying “goalstructure” using rich (and enrichable) con-straint descriptions that include the activities’intended impact on the environment. Thisallows for more precise and useful monitoringof plan execution, allowing plans to be adjustedor repaired as circumstances change. It canexploit information about the dynamicallychanging context and status of the agents andproducts involved (for example, throughemerging geolocation services for people andproducts). It also provides for the real-timecommunication of activities and tasks betweenhuman and automated resources.

I-X agents and the underpinning <I-N-C-A> ontology can be used in a range of systems

including supportive interfaces for humansand organizations and potentially in intelli-gent sensors and robotic devices. The I-Xagents and <I-N-C-A> ontology can thus actas a shared mechanism for coordinating thesesystems, sensors, and devices and giving themintelligent planning and process support (seefigure 3).

I-Rescue and I-X systems aim to be partof a future environment that has

• multilevel emergency response and aidsystems;

• a backbone for progressively more com-prehensive aid and emergency response;

• applications for aid-orientated commer-cial services;

• a robust, secure, resilient, and distributedsystem of systems;

• advanced knowledge and collaborationtechnologies;

• low-cost pervasive sensors, computing,and communications; and

• changes in building codes, regulations,and practices.

Advanced KnowledgeTechnologies e-Response

The Collaborative Advanced KnowledgeTechnologies in the Grid (CoAKTinG) projectused technologies from the AKT program tosupport distributed scientific collaboration inan emergency response situation—in particu-

lar, a scenario involving managing an oil spill.6

Focusing on the interchange between humansin the scenario, CoAKTinG provided tools toassist scientific collaboration by integratingintelligent meeting spaces, ontologically anno-tated media streams from online meetings,decision rationale and group memory capture,meeting facilitation, planning and coordinationsupport, argumentation, and instant messagingand presence.

AKT focuses on the provision of next-gen-eration knowledge technologies, particularlyin the context of the Semantic Web as both amedium and a target domain for these tech-nologies. Work on the AKT project focuseson a challenge problem dealing with the after-math of a civil cargo aircraft crashing into alarge city—an actual scenario for which con-tingency plans are in place. This research looksat how to use the Semantic Web to help makesense of the situation, both to guide emergencyresponders and to find appropriate specializedrescue and medical capabilities.

OpenKnowledge e-ResponseOpenKnowledge is a European Union pro-

ject involving the Edinburgh, Southampton,and the Open Universities in the UK, alongwith Amsterdam, Trento, and Barcelona. Theproject aims to provide an open frameworkfor agent interaction and coordination inknowledge-rich environments. It focuses ontwo challenge problems, one of which is

FireGrid

Integration of HPC-implemented simulations with sensors and C/C (command and control)

Grid-enabled, sensor-steeredcoupled “ensemble” HPC simulations

Coupled HPC simulationsintegrated with sensor data

High-performance computing-implemented simulations integrated with C/C

Grid-enabled, sensor- steered computations

HPC-implemented coupled andstand-alone simulation codes

Simulation output filteringand C/C as a Grid service

Sensor computationintegration

Communicationsand networking

Sensors andsystems

Gridtechnologies

HPCtechnologies

Loosely coupled simulation codes andsemianalytical models for extrapolation

Egresssoftware

Structuressoftware

CFDsoftware

Commandand control

Grid-enabled sensorcommunications

Figure 2. Integration of FireGrid technologies.3

emergency monitoring and management.The project chose this domain as a test bedbecause it demands a combination of geo-graphical and geopresence knowledge along-side active support for collaboration andplanning in multiagent, dynamic situations.

The need to harness electronic communi-cation networks in emergency situations hasbeen recognized as a European research pri-ority. The EU Emergency Response Grid pro-gram (www.cordis.lu/ist/grids/emergency_response_grid.htm) says

In times of crisis—be it a natural disaster, ter-rorist attack, or infrastructure failure—mobileworkers need to work together in time-criticaland dangerous situations. Real-time access toinformation and knowledge, powered by Grids,will help save lives. Crises are complex situa-tions, with large numbers and varieties of mobileworkers—medical and rescue teams, police, firefighters, and other security personnel—appear-ing on the spot at short notice. These differentteams come from different organizations, andgenerally have incomplete or even contradictoryknowledge of the crisis situation.

Emergency response coalitions are oftenhighly dynamic and opportunistic “commu-

nities of practice.” In this context, the Open-Knowledge e-Response test bed involves theexploration of the following peer-to-peerservices:

• Network data sources. Sensor data flowsmust be coordinated in the large. Thesedata will be classified locally and must bemade available to different “experts orpeers” on the basis of contextual classifi-cation schema.

• Collaborative services. These supportplanning, communication, and coordina-tion within the expert peers community.

• Mitigation assessment services. As poten-tial emergency situations are identified,mitigation needs can be determined andprioritized.

• Preparedness services. Such services sup-port those activities that prepare for actualemergencies.

Collaborative operations forpersonnel recovery

Personnel recovery (search and rescue)teams must operate under intense pressure, tak-

ing into account not only hard logistics but also“messy” factors such as a decision’s social orpolitical implications. The Collaborative Oper-ations for Personnel Recovery (Co-OPR) pro-ject has developed decision support for sense-making in such scenarios, seeking to exploithuman and machine reasoning’s complemen-tary strengths. Co-OPR integrates the Com-pendium sense-making support tool for real-time information and argument mapping7 withthe I-X artificial intelligence planning and exe-cution framework8 to support group activityand collaboration. Both share a commonmodel for dealing with issues, the refinementof options for the activities to be performed,handling constraints, and recording other infor-mation. The tools span a spectrum from Com-pendium (which is very flexible, with few con-straints on terminology and content) to I-X(with its knowledge-based reliance on richdomain models and formal conceptual mod-els, or ontologies). In a simulated exerciseinvolving personnel recovery during a UNpeacekeeping operation, with roles played bymilitary planning staff, external evaluatorsjudged the Co-OPR tools to be very effective.

T h e F u t u r e o f A I

60 www.computer.org/intelligent IEEE INTELLIGENT SYSTEMS

Figure 3. An I-X Process Panel and its accompanying tools, engaged in coordinating the response to a simulated environmentalemergency.

MAY/JUNE 2006 www.computer.org/intelligent 61

International systems andorganizations

The infrastructure and organizations thatwould be required to make the vision of thehelpful environment possible are being put inplace. The Galileo European Satellite Naviga-tion System (http://europa.eu.int/comm/dgs/energy_transport/galileo) and its mobile geolo-cation and emergency response services pro-gram will be another to spur development. Bothcommercial and freely provided emergencyresponse facilities are being interwoven toensure active development and support over along period.

Examples of the type of “helpful organiza-tion” that will enact this vision are already start-ing to emerge. The Multinational PlanningAugmentation Team (www2.apan-info.net/mpat), an organization consisting of 33 PacificRim nations, has been developing sharedknowledge and procedures to assist inresponses to regional crises.9 MPAT used com-puter collaboration aids and a simple broker-ing system during the December 2004 to Feb-ruary 2005 Indian Ocean tsunami response tohelp affected countries access the specializedcapabilities of response organizations moreeffectively. MPAT is an excellent example ofpeople training and working together to ensureemergency response readiness. MPAT wouldbe a prime beneficiary and exploiter of anyfuture helpful environment.

On a smaller scale, the Washington, D.C.area is conducting the Capital Wireless Inte-grated Network (http://capwin.org), led byIBM and the University of Maryland. Thisprogram works with local services to supportresponses by all agencies to incidents occur-ring on a stretch of freeway in the Washing-ton, D.C. area. It aims to show coordinatedwireless-computing services’potential valuefor integrated and cohesive response acrosspolice units, ambulance services, emergencysupport teams, fire brigades, hazardous-material units, and the military.

In the not-too-distant future, every citizen,vehicle, in-transit package, and other

“active device” might be treated as a poten-tial sensor or responder. Individuals or vehi-cles that need help, as well as local, regional,national, and international emergency agen-cies, could look up specialized capabilitiesor find local assistance through a much moreresponsive and effective environment. Sys-tems could interoperate to enable preventa-

tive measures to be taken so that those inimminent danger could be forewarned bytheir own systems, and by the people, vehi-cles, and buildings around them.

These would support a diverse range ofuses, such as

• disaster response and evacuation,• terrorism incident response,• civil accidents,• disease control,• business continuity,• family emergencies,• transportation aids,• help desks, and• procedural assistance.

We could create a truly helpful environ-ment that everyone can access. In the second50 years of AI, let’s make it happen!

AcknowledgmentsThis article is based on evidence given to US and

UK government agencies concerned with responsesto emergency situations and with “Grand Challenges”for computer science research. An earlier version waspresented at a public forum at the Royal Museum ofScotland in February 2005 on responses by the sci-entific community to the Indian Ocean tsunami.

Projects mentioned in this article have beensponsored by a number of organizations. The Uni-versity of Edinburgh and research sponsors areauthorized to reproduce and distribute reprints andonline copies for their purposes notwithstandingany copyright annotation hereon. The views andconclusions contained herein are the author’s andshouldn’t be interpreted as necessarily represent-ing the official policies or endorsements, eitherexpressed or implied, of other parties.

References

1. Safety.net, “Ubiquitous Computing for Disas-ter Mitigation, Response, and Recovery,” pre-sentation at Computing Research Assn. Conf.Grand Research Challenges in Computer Sci-ence and Eng., July 2002; www.cra.org/Activ-ities/grand.challenges/slides/ubiquitous.pdf.

2. H. Kitano and S. Tadokoro, “RoboCup Res-cue: A Grand Challenge for Multiagent andIntelligent Systems,” AI Magazine, vol. 22,no. 1, 2001, pp. 39–51.

3. D. Berry, A. Usmani, J. Terero, A. Tate, S.McLaughlin, S. Potter,A. Trew, R. Baxter, M.Bull, and M. Atkinson, “FireGrid: IntegratedEmergency Response and Fire Safety Engi-neering for the Future Built Environment,”Proceedings of UK e-Science Programme AllHands Meeting (AHM 05), UK e-ScienceProgramme, Sep. 2005; www.aiai.ed.ac.uk/project/ix/documents/2005/2005-escience-ahm-firegrid.doc.

4. A. Tate et al., “Intelligent Agents for CoalitionSearch and Rescue Task Support,” Proc. 19thNat’l Conf. Am. Assoc. Artificial Intelligence(AAAI 04), 2004,AAAI Press, pp. 1038–1039;www.aiai.ed.ac.uk/project/ix/documents/2004/2004-aaai-isd-tate-cosarts.pdf.

5. A. Tate, “<I-N-C-A>: An Ontology forMixed-Initiative Synthesis Tasks,” Proc.Workshop Mixed-Initiative Intelligent Sys-tems, Int’l Joint Conf. Artificial Intelligence(IJCAI 03), 2003.

6. S. Buckingham Shum, D. De Roure, M. Eisen-stadt, N. Shadbolt, and A. Tate, “CoAKTinG:Collaborative Advanced Knowledge Technolo-gies in the Grid,” presented at 2nd WorkshopAdvanced Collaborative Environments, 11thIEEE Int’l Symp. High Performance DistributedComputing (HPDC 11), 2002, IEEE CS Press;www.aiai.ed.ac.uk/project/ix/documents/2002/2002-wace-coakting.pdf.

7. S. Buckingham Shum et al., “HypermediaSupport for Argumentation-Based Rationale:15 Years On from gIBIS and QOC,” Ratio-nale Management in Software Engineering,Springer, 2006, pp. 111–132.

8. A. Tate, J. Dalton, and J. Stader, “I-P2—Intel-ligent Process Panels to Support CoalitionOperations,” Proc. 2nd Int’l Conf. KnowledgeSystems for Coalition Operations (KSCO 02),2002, pp. 184–190; www.aiai.ed.ac.uk/project/ksco/ksco-2002.html.

9. S.A. Weide, “Multinational Crisis Response inthe Asia-Pacific Region:The Multinational Plan-ning Augmentation Team Model,” The Liaison,Feb. 2006; www.coe-dmha.org/liaison.htm.

T h e A u t h o rAustin Tate is thedirector of the Artifi-cial Intelligence Appli-cations Institute andthe personal chair ofknowledge-based sys-tems at the Universityof Edinburgh. He is thechief technical officer

of I-C2 Systems, a company seeking to developadvanced aids for emergency response. Hisinternationally sponsored research workinvolves advanced knowledge and planningtechnologies, especially for use in emergencyresponse. He received his PhD in machine intel-ligence from the University of Edinburgh. He’sa fellow of the Royal Society of Edinburgh(Scotland’s National Academy) and severalother bodies and an advisory board member forIEEE Intelligent Systems. Contact him at AIAI,Univ. of Edinburgh, Appleton Tower, CrichtonSt., Edinburgh EH8 9LE, UK; a.tate@ed.ac.uk.