Daniel D. Corkill: On-Line Publications
This web page is provided for convenience in accessing on-line
versions of selected publications. Please observe that the copyright
terms on some publications restrict redistribution or non-personal use
without written permission of the copyright holder.
- Leveraging
Failures to Enhance Hierarchical Concept Learning when Training and
Testing are Limited, Huzaifa Zafar and Daniel D. Corkill. Technical
Report UM-CS-2011-026, Department of Computer Science, University of
Massachusetts Amherst, July 2011.
Hierarchical concept learning constructs higher-level concepts using
previously learned prerequisite concepts. In our DARPA Bootstrapped Learning
program work, we faced an especially challenging context where only a small
number of training instances for each concept were provided to the learning
system. Such limited instruction forces even the most skillful learner to make
assumptions about the concept being taught—assumptions that can be
incorrect. Given this uncertainty, multiple candidates may be proposed by
learning algorithms for the concept, each stemming from different assumptions
that are consistent with the training. We developed a novel control strategy
for managing the use of hypothesized concept candidates in higher-level
learning. This Concept Candidate Management (CCM) strategy is based on three
key ideas: 1) limiting prerequisite-candidate combinatorics by operating with
a single selected candidate for each concept at any time, 2) using learning
failure to select a different candidate for a direct or indirect prerequisite
concept, and 3) using differences observed as candidates are used to guide
candidate selection. We evaluated CCM in MABLE, an electronic student that
performs bootstrapped concept learning. Using the CCM strategy, MABLE learned
concepts that were not successfully learned otherwise—without any
additional training or testing and without any changes to learning algorithms.
(PDF)
- Deploying
Power-Aware, Wireless Sensor Agents, Daniel D. Corkill. The Computer
Journal, 54(8):392–405, March 2011.
Developing sensor agents that can be deployed untethered in the field
presents significant challenges in adapting to hardware, communication,
power, and environmental limitations. Real-world characteristics dictate
agent behavior and operating strategies, sometimes quite differently from
often held assumptions and intuitions. In this article, we describe the
sensor-agent hardware and blackboard-system software used in CNAS
(Collaborative Network for Atmospheric Sensing), an agent-based, power-aware
sensor network for ground-level atmospheric monitoring. CNAS is
representative of a class of battery-powered, wireless sensor networks in
which the distance separating deployed sensor agents is near the limit of
their WiFi communication range. To conserve battery power, CNAS sensor
agents must have their wireless radios turned off most of the time, as even
having them turned on consumes significant power. We discuss how CNAS
agents collaborate using only periodic radio availability and consider how
different hardware and communication capabilities would change CNAS
strategies. We also relate challenges that had to be addressed during
deployments of CNAS at military exercises held in the summer heat in
Wisconsin and in the rain and mud in Queensland, Australia. We conclude
with research on improving CNAS responsiveness with limited radio
availability and on potential next-generation CNAS hardware. (PDF)
- Reducing
Online Model Development Time by Agents using Constraints between Shared
Observations, Huzaifa Zafar and Daniel D. Corkill. The Computer
Journal, 53(8):1302–1314, October 2010.
A situated agent must determine aspects of its environment in order to make
appropriate decisions. This determination must be done quickly, as performance
can suffer until each agent develops a sufficiently accurate model of its
environment. We introduce a two phase model-development approach that leads to
a significant reduction in the online (post-deployment) time required to
determine environmental models. During the pre-deployment phase, an
incompletely specified, site-independent model of an agent's environment is
developed, with the site-dependent features represented as parameters in the
model. This pre-deployment model is then completed during the post-deployment
phase by determining the model parameters using constraints between local and
shared observations. In this article, we use this approach in developing an
environmental model for potential solar visibility and panel collection
characteristics by each agent in a power-aware wireless sensor network. We
show that, by using temporal and spatial constraints between shared
observations, each agents can complete its solar-harvesting model using only
the first and second day observations as compared to 10 days of observations
required by the power-management algorithm of Kansal et al. (PDF)
- Agent
Technologies for Sensor Networks, Alex Rogers, Daniel D. Corkill, and
Nicholas R. Jennings. IEEE Intelligent Systems,
24(2):13–17, March/April 2009.
Wireless sensor networks are increasingly seen as a solution to the problem
of performing continuous wide-area monitoring in many environmental, security,
and military scenarios. Such networks consist of small, battery-powered
devices that are physically distributed over a wide area and connected through
a wireless communication network. Since these networks often must collect data
over extended periods of time and are deployed in inhospitable environments
where replacing batteries is inconvenient or impossible, much of the research
in this domain addresses the challenge of minimizing each sensor's energy
needs. To this end, researchers have developed a wide range of
energy-efficient sensor nodes and wireless communication protocols and
demonstrated them in varied applications. This article describes three example
applications. In each case, researchers have demonstrated the work in the
wild, implemented it on real sensor hardware, deployed it in real, hostile
environments, and evaluated it on real sensor data. (PDF)
- Reporting Down Under: A CNAS (Collaborative
Network for Atmospheric Sensing) Update, Daniel D. Corkill. In
Second International Workshop on Agent Technology for Sensor
Networks (ATSN-08), Estoril, Portugal, pages 25–32, May 2008.
We briefly review the sensor-agent hardware and blackboard-system software
used in CNAS (Collaborative Network for Atmospheric Sensing), an
agent-based, power-aware sensor network for ground-level atmospheric
monitoring. We then describe experiences and lessons learned from field
deployments of CNAS at the 2007 Talisman-Saber Combined Exercise held in
Queensland, Australia. We conclude with an overview of CNAS research
performed since Talisman-Saber that focuses on: 1) improving CNAS
performance and responsiveness with limited radio availability, 2)
power-aware reasoning associated with solar harvesting obtained from a
rollable solar panel at each sensor agent, and 3) potential next-generation
CNAS hardware. (PDF)
- Determining Confidence When
Integrating Contributions from Multiple Agents, Raphen
Becker and Daniel D. Corkill. In Sixth International
Joint Conference on Autonomous Agents and Multi-Agent
Systems (AAMAS 2007), Honolulu, Hawaii, pages
449–456, May 2007.
Integrating contributions received from other agents is an
essential activity in multi-agent systems (MASs). Not only must
related contributions be integrated together, but the confidence in
each integrated contribution must be determined. In this paper we
look specifically at the issue of confidence determination and its
effect on developing “principled,” highly collaborating
MASs. A domain-independent analysis model is presented that can be
used to measure the sensitivity of a collaborative problem-solving
system to potentially incorrect confidence-integration assumptions.
The analysis model can be used to determine confidence bounds on
integrated contributions and to identify where efforts to improve
contribution-dependency estimates lead to the greatest improvement
in solution-confidence accuracy. (PDF)
- Turn Off Your Radios! Environmental
Monitoring Using Power-Constrained Sensor Agents,
Daniel D. Corkill, Douglas Holzhauer, and Walter Koziarz.
In First International Workshop on Agent Technology for
Sensor Networks (ATSN-07), Honolulu, Hawaii, pages
31–38, May 2007.
CNAS (Collaborative Network for Atmospheric Sensing) is an
agent-based, power-aware sensor network for ground-level atmospheric
monitoring. In many multi-agent applications, reducing message
transmission is a primary objective. In CNAS, however, it's not the
cost of sending messages, but when messages can be sent
that is the driving communication constraint. CNAS agents must have
their radios turned off most of the time, as even listening
consumes significant power. CNAS requires agent policies that can
intelligently meet operational requirements while communicating only
during intermittent, mutually established, communication windows.
CNAS agents and their hardware and blackboard-system software
architectures are described, as well as experiences and lessons
learned from a field deployment of CNAS at the 2006 PATRIOT Exercise
held in July 2006 at Fort McCoy, Wisconsin. (PDF)
- Representation and
Contribution-Integration Challenges in Collaborative Situation
Assessment, Daniel D. Corkill. In Proceedings of the Eighth
International Conference on Information Fusion (Fusion 2005),
Philadelphia, Pennsylvania, pages xxix–xxxi, July 2005.
Blackboard systems are an ideal architecture for situation
assessment involving large data volumes and heterogeneous data and
knowledge sources. However, the ad hoc confidence and belief values
used in traditional blackboard applications has led to criticism of
the blackboard approach and spawned efforts to combine collaborative
blackboard-system techniques with more “principled”
graphical-network representations. Two important
collaborative-assessment challenge areas are discussed in this brief
position paper: 1) principled blackboard representations and 2)
principled integration of contributions made by independent
knowledge-source entities. The complexity of these challenges is
highlighted using a very simple assessment scenario. (PDF)
- Collaborating Software: Blackboard and
Multi-Agent Systems & the Future, Daniel D. Corkill. In
Proceedings of the International Lisp Conference, New York, New
York, October 2003.
AI researchers have used the paradigm of collaborating software
systems to tackle large and difficult problems. This invited
presentation compares and contrasts two markedly different
collaborating-software approaches: blackboard systems and
multi-agent systems. Examining collaborating software from both
perspectives provides insights into the nature of collaboration,
reveals unresolved problems in integrating disparate contributions,
and underscores issues in coordinating collaborative activities.
(PDF)
- Mixed-Initiative Management of Dynamic
Business Processes, Zachary B. Rubinstein and Daniel
D. Corkill. Proceedings of the 2003 IEEE International Workshop on
Soft Computing in Industrial Applications, pages 39–44, Binghamton,
New York, June 2003.
Describes the ProME process-management environment, focusing on how
human process managers and participants interact with a dynamic,
on-line model of executing dynamic processes to proactively manage
and operate in dynamic business processes. Having the best
information available about a process and its future provides
managers with the time needed to detect and understand impending
process anomalies and to develop and implement effective
interventions. Furthermore, enabling managers to update the
executing process representation and having the ProME environment
push the effects of those modifications to the relevant participants
reduces the time it takes to implement remedies. ProME was used in
a commercial product for managing design processes in the automotive
and aerospace industries. (PDF)
- Live-Representation Process Management,
Daniel D. Corkill, Zachary B. Rubinstein, Susan E. Lander, and Victor
R. Lesser. Proceedings of the 5th International Conference on
Enterprise Information Systems, pages 203–208, Angers, France,
April 2003.
Describes the "live-representation" approach to managing and
working in complex, dynamic business processes. In this promising new
approach, important aspects of business-process modeling, project
planning, project management, resource scheduling, process automation,
execution, and reporting are integrated into a detailed, on-line
representation of planned and executing processes. (PDF)
- When Workflow Doesn't Work: Issues in
managing dynamic processes, Daniel D. Corkill. Proceedings of
the Design Project Support using Process Models Workshop, Sixth
International Conference on Artificial Intelligence in Design,
Worcester, Massachusetts, June 2000.
Explains why traditional process-execution systems cannot address
the management requirements of dynamic design processes and describes
a new type of decision-support software developed specifically to
address the management of dynamic design processes. (PDF)
- Diversity in Agent
Organizations, Daniel D. Corkill and Susan E. Lander.
A concise and accessible presentation of the issues associated with
agent-based organizations. This is the full version of the article
"Agent Organizations" that was published in Object Magazine,
8(4)41-47, May 1998. (HTML)
- Countdown to Success: Dynamic objects,
GBB, and RADARSAT-1, Daniel D. Corkill. Communications of the
ACM, 40(5):848-858, May 1997.
An invited account of the importance that blackboard-system and
dynamic-object capabilities played in the rapid development of the
ground-based portion of the RADARSAT-1 Mission Control
System. (PDF)
- Designing Integrated Engineering
Environments: Blackboard-Based Integration of Design and Analysis
Tools, Susan E. Lander, Scott M. Staley, and Daniel
D. Corkill. Concurrent Engineering: Research and Applications,
Special Issue on the Application of Multi-Agent Systems to Concurrent
Engineering, 4(1):59-72, March 1996.
Describes the use of blackboard technology in creating the
agent-based RRM integrated concurrent-engineering environment for
automotive design at Ford. (PostScript)
- Blackboard Systems, Daniel D. Corkill.
An introduction to blackboard systems. This article discusses the
characteristics and potential of blackboard systems. It describes what
a blackboard system is and what it is not, considerations for using
blackboard systems, and how to get started. This is the unabridged
version of the article that appeared in AI Expert 6(9):40-47,
September 1991. (PDF)
- Embedable Problem-Solving Architectures: A
Study of Integrating OPS5 with UMass GBB, Daniel
D. Corkill. IEEE Transactions on Knowledge and Data Engineering
3(1):18-24, March 1991.
Discusses the issues involved in creating a problem-solving
architecture that can be tightly embedded within other architectures
and coexist with multiple instances of itself and of other
problem-solvers. An detailed example describing modifications and
enhancements made to the public-domain version of OPS5 in order to
embed it as an integral KS language within the UMass GBB system is
presented. (PDF)
- Design Alternatives for
Parallel and Distributed Blackboard Systems, Daniel
D. Corkill. In V. Jagannathan, Rajendra Dodhiawala, and
Lawrence S. Baum, editors, Blackboard Architectures and
Applications, pages 99–136. Academic Press, 1989.
This book chapter discusses important issues in the design of parallel
and distributed blackboard architectures, focusing on issues of
performance and the maintenance of semantic consistency of the
blackboard. (PDF)
- Trends in Cooperative Distributed Problem
Solving, Edmund H. Durfee, Victor R. Lesser, and Daniel D. Corkill.
IEEE Transactions on Knowledge and Data Engineering, 1(1):63-82,
March, 1989 (Invited paper).
A invited survey of Cooperative Distributed Problem Solving circa
1989. (PostScript)
- Achieving Flexibility, Efficiency, and
Generality in Blackboard Architectures, Daniel D. Corkill, Kevin
Q. Gallagher, and Philip M. Johnson. In Proceedings of the National
Conference on Artificial Intelligence (AAAI-87), pages 18–23,
Seattle, Washington, July 1987. (Also published in Readings in
Distributed Artificial Intelligence, Alan H. Bond and Les Gasser,
editors, pages 451–456, Morgan Kaufmann, 1988.)
A discussion of the high-performance, dimensional data-abstraction
techniques developed for the UMass GBB blackboard-development
system. (PDF)
- Use of Meta-Level Control for Coordination
in a Distributed Problem-Solving Network, Daniel D. Corkill and
Victor R. Lesser. Proceedings of the Eighth International Joint
Conference on Artificial Intelligence (IJCAI-83), pages 748–756,
Karlsruhe, Federal Republic of Germany, August 1983.
This summary of Corkill's dissertation research describes the first
research to specifically explore the use of organizational self-design
and coordination in multi-agent systems. Also discussed is the
importance of agent skepticism where agents actively monitor
the internal conflict between organizationally specified activities
and locally desirable actions. An implementation of these ideas is
briefly described along with the results of preliminary experiments
with various network problem-solving strategies specified via
organizational structuring. (PDF)
- Unifying Data-Directed and Goal-Directed Control:
An example and experiments, Daniel D. Corkill, Victor R. Lesser,
and Eva Hudlicka. Proceedings of the National
Conference on Artificial Intelligence (AAAI-82), pages 143-147,
Pittsburgh, Pennsylvania, August 1982.
A description of a blackboard architecture that integrates
data-directed and goal-directed control into a single, uniform
framework. This architecture was used in the UMass Distributed
Vehicle Monitoring Testbed (PDF)
- Functionally Accurate, Cooperative
Distributed Systems, Victor R. Lesser and Daniel D. Corkill.
IEEE Transactions on Systems, Man, and Cybernetics,
SMC-11(1):81-96, January 1981.
Original presentation of the idea that knowledge-based AI
techniques could be used to design distributed systems that could
operate with local views that were not complete, consistent, and
up to date. (PDF)
- Hierarchical Planning in a Distributed
Problem-Solving Environment, Daniel D. Corkill.
Describes extensions to Sacerdoti's NOAH hierarchical planning
system for use with multiple, distributed planning agents. This the
full version of the paper that appeared in the Proceedings of the
Sixth International Joint Conference on Artificial Intelligence,
pages 168-175, Tokyo, August 1979. (HTML)
- Curriculum Vitae (PDF)
|