The primary topics for the SYROCO 2009 will be "Robot Control in Human-
Robot Dynamic Interaction". Robot control technology is widely used for
space, surgery, rehabilitation, micro machine, entertainment, underwater,
civil engineering etc. It will continue to play an increasing role in the
area of human-robot dynamic interaction technology. The SYROCO 2009 will
also try to cover the whole range of areas in the field of robot control
including mobile robots. Contributions on basic research as well as on
relevant applications are included.
See the CFP.
Thursday, January 29, 2009
Wednesday, January 28, 2009
Victor Lesser Distinguished Dissertation Award
Every year IFAAMAS gives out an award for the best agents related dissertation. In 2008, the award went to Radu Jurca for the dissertation entitled "Truthful Reputation Mechanisms for Online Systems". Radu went to Google (Switzerland) after finishing. Vincent Conitzer won it the year before for a thesis entitled " Computational Aspects of Preference Aggregation." Vince is now a tenure track Professor at Duke in the U.S.. (This information took some serious googling to find out, which it shouldn't have! IFAAMAS get your act together and publicize this award.)
Victor Lesser is considered the Great Grandfather of the agents field -- Sandip Sen's students are academic great grandchildren of Prof. Lesser. It won't be long until Lesser is a Great Great Grandfather (see http://aigp.eecs.umich.edu/)
To submit nominations for this years award:
Nominations are invited for the 2008 Victor Lesser Distinguished Dissertation Award sponsored by IFAAMAS, the International Foundation for Autonomous Agents and Multiagent Systems (http://www.ifaamas.org).
This award includes a certificate signed by the IFAAMAS Chair and 1500EUR. Eligible doctoral dissertations are those defended between January 1, 2008 and December 31, 2008 in the area of Autonomous Agents or Multiagent Systems.
For more detail see here.
Victor Lesser is considered the Great Grandfather of the agents field -- Sandip Sen's students are academic great grandchildren of Prof. Lesser. It won't be long until Lesser is a Great Great Grandfather (see http://aigp.eecs.umich.edu/)
To submit nominations for this years award:
Nominations are invited for the 2008 Victor Lesser Distinguished Dissertation Award sponsored by IFAAMAS, the International Foundation for Autonomous Agents and Multiagent Systems (http://www.ifaamas.org).
This award includes a certificate signed by the IFAAMAS Chair and 1500EUR. Eligible doctoral dissertations are those defended between January 1, 2008 and December 31, 2008 in the area of Autonomous Agents or Multiagent Systems.
For more detail see here.
Friday, January 23, 2009
A Multiagent Reinforcement Learning Algorithm with Non-linear Dynamics
A couple of newly published JAIR articles are of potential interest to MAS researchers, this one in particular:
S. Abdallah and V. Lesser (2008)
"A Multiagent Reinforcement Learning Algorithm with Non-linear Dynamics",
Volume 33, pages 521-549
For quick access go to here.
Abstract:
Several multiagent reinforcement learning (MARL) algorithms have been proposed to optimize agents' decisions. Due to the complexity of the problem, the majority of the previously developed MARL algorithms assumed agents either had some knowledge of the underlying game (such as Nash equilibria) and/or observed other agents actions and the rewards they received.
We introduce a new MARL algorithm called the Weighted Policy Learner (WPL), which allows agents to reach a Nash Equilibrium (NE) in benchmark 2-player-2-action games with minimum knowledge. Using WPL, the only feedback an agent needs is its own local reward (the agent does not observe other agents actions or rewards). Furthermore, WPL does not assume that agents know the underlying game or the corresponding Nash Equilibrium a priori. We experimentally show that our algorithm converges in benchmark two-player-two-action games. We also show that our algorithm converges in the challenging Shapley's game where previous MARL algorithms failed to converge without knowing the underlying game or the NE. Furthermore, we show that WPL outperforms the state-of-the-art algorithms in a more realistic setting of 100 agents interacting and learning concurrently.
An important aspect of understanding the behavior of a MARL algorithm is analyzing the dynamics of the algorithm: how the policies of multiple learning agents evolve over time as agents interact with one another. Such an analysis not only verifies whether agents using a given MARL algorithm will eventually converge, but also reveals the behavior of the MARL algorithm prior to convergence. We analyze our algorithm in two-player-two-action games and show that symbolically proving WPL's convergence is difficult, because of the non-linear nature of WPL's dynamics, unlike previous MARL algorithms that had either linear or piece-wise-linear dynamics. Instead, we numerically solve WPL's dynamics differential equations and compare the solution to the dynamics of previous MARL algorithms.
S. Abdallah and V. Lesser (2008)
"A Multiagent Reinforcement Learning Algorithm with Non-linear Dynamics",
Volume 33, pages 521-549
For quick access go to here.
Abstract:
Several multiagent reinforcement learning (MARL) algorithms have been proposed to optimize agents' decisions. Due to the complexity of the problem, the majority of the previously developed MARL algorithms assumed agents either had some knowledge of the underlying game (such as Nash equilibria) and/or observed other agents actions and the rewards they received.
We introduce a new MARL algorithm called the Weighted Policy Learner (WPL), which allows agents to reach a Nash Equilibrium (NE) in benchmark 2-player-2-action games with minimum knowledge. Using WPL, the only feedback an agent needs is its own local reward (the agent does not observe other agents actions or rewards). Furthermore, WPL does not assume that agents know the underlying game or the corresponding Nash Equilibrium a priori. We experimentally show that our algorithm converges in benchmark two-player-two-action games. We also show that our algorithm converges in the challenging Shapley's game where previous MARL algorithms failed to converge without knowing the underlying game or the NE. Furthermore, we show that WPL outperforms the state-of-the-art algorithms in a more realistic setting of 100 agents interacting and learning concurrently.
An important aspect of understanding the behavior of a MARL algorithm is analyzing the dynamics of the algorithm: how the policies of multiple learning agents evolve over time as agents interact with one another. Such an analysis not only verifies whether agents using a given MARL algorithm will eventually converge, but also reveals the behavior of the MARL algorithm prior to convergence. We analyze our algorithm in two-player-two-action games and show that symbolically proving WPL's convergence is difficult, because of the non-linear nature of WPL's dynamics, unlike previous MARL algorithms that had either linear or piece-wise-linear dynamics. Instead, we numerically solve WPL's dynamics differential equations and compare the solution to the dynamics of previous MARL algorithms.
Thursday, January 22, 2009
Smart Dust
"Molecular Forklifts Overcome Obstacle To 'Smart Dust':
Chip-based labs have been developed in recent years as portable tools to gauge the presence of bioweapons, pollution, or to conduct on-the-spot blood tests. They are essentially assays, or ways to test for different pathogens, chemicals or compounds. The research was funded primarily by the Defense Advanced Research Projects Agency, with additional support from the Office of Naval Research"
Read article.
The dust is unlikely to run any version of Java in the near term.
Chip-based labs have been developed in recent years as portable tools to gauge the presence of bioweapons, pollution, or to conduct on-the-spot blood tests. They are essentially assays, or ways to test for different pathogens, chemicals or compounds. The research was funded primarily by the Defense Advanced Research Projects Agency, with additional support from the Office of Naval Research"
Read article.
The dust is unlikely to run any version of Java in the near term.
Video wrap of commercial advances in 2008
Quick video wrap up of robots from companies like Honda, iRobot and Toyota that debuted in 2008, from PC World.
Watch the video (after enduring the ad.)
The physical capabilities of these things are pretty impressive, but, likely because the video all comes from big demos, the movement is slow and careful. No doubt, in the comfort of the lab, the behavior is a lot smoother and faster.
Watch the video (after enduring the ad.)
The physical capabilities of these things are pretty impressive, but, likely because the video all comes from big demos, the movement is slow and careful. No doubt, in the comfort of the lab, the behavior is a lot smoother and faster.
Wednesday, January 21, 2009
ICRA versus AAMAS acceptance rates
Although down from its early very high acceptance rates, the premier robotics conference, ICRA, accepts about 40% of submitted papers (43% in 2009). This is nearly double the AAMAS acceptance rate and about double the acceptance rate of other major AI conferences (IJCAI is in the teens).
ICRA is a large conference, much bigger than AAMAS and similar in size to IJCAI (the Grandaddy of AI conferences.) It has more than five times the number of presentations than AAMAS and three times as many oral presentations as IJCAI.
There are clearly different philosophies of the different organizing committees: ICRA accepts more papers, has a bigger conference, AAMAS and IJCAI accept less papers, limiting the conference size (since, overwhelmingly, relatively few people attend a conference where they do not have a paper).
The question is which is a better approach for advancing science (admittedly, only one of several reasons for conferences).
A bigger conference lets more people get together and discuss, possibly advancing science faster than the smaller conferences. The counter argument is that the smaller conferences are higher quality and therefore advance science more efficiently.
From a game theory perspective, I would imagine that the more selective conferences have even higher relative quality than the acceptance rates suggest, because authors only submit high quality work.
Perhaps both work for their respective fields. Robotics is a slightly younger field, with lots of ideas to be vetted, while AI is general is a bit more mature. The big robotics conference lets lots of ideas in, the smaller AI conferences efficiently cut out some of the noise.
I am not sure there is a right or wrong answer.
ICRA is a large conference, much bigger than AAMAS and similar in size to IJCAI (the Grandaddy of AI conferences.) It has more than five times the number of presentations than AAMAS and three times as many oral presentations as IJCAI.
There are clearly different philosophies of the different organizing committees: ICRA accepts more papers, has a bigger conference, AAMAS and IJCAI accept less papers, limiting the conference size (since, overwhelmingly, relatively few people attend a conference where they do not have a paper).
The question is which is a better approach for advancing science (admittedly, only one of several reasons for conferences).
A bigger conference lets more people get together and discuss, possibly advancing science faster than the smaller conferences. The counter argument is that the smaller conferences are higher quality and therefore advance science more efficiently.
From a game theory perspective, I would imagine that the more selective conferences have even higher relative quality than the acceptance rates suggest, because authors only submit high quality work.
Perhaps both work for their respective fields. Robotics is a slightly younger field, with lots of ideas to be vetted, while AI is general is a bit more mature. The big robotics conference lets lots of ideas in, the smaller AI conferences efficiently cut out some of the noise.
I am not sure there is a right or wrong answer.
Monday, January 19, 2009
Number of AAMAS Workshops
I was surprised by the number of workshops proposed for AAMAS'09, so I went back and look at the 6 previous conferences.
2009: 27 workshops
2008: 26 workshops, with 3 more cancelled.
2007: 31 workshops, with 1 more cancelled.
2006: 28 workshops, with 3 more cancelled.
2005: 28 workshops, with 1 more cancelled.
2004: 24 workshops, with 1 more cancelled.
2003: 22 workshops.
2002: 17 workshops.
So, I guess this year is pretty much in line with previous years. After a quick ramp up, there has been between 24-30 for the last 6 years. I wonder which 1.8 will get cancelled this year?
2009: 27 workshops
2008: 26 workshops, with 3 more cancelled.
2007: 31 workshops, with 1 more cancelled.
2006: 28 workshops, with 3 more cancelled.
2005: 28 workshops, with 1 more cancelled.
2004: 24 workshops, with 1 more cancelled.
2003: 22 workshops.
2002: 17 workshops.
So, I guess this year is pretty much in line with previous years. After a quick ramp up, there has been between 24-30 for the last 6 years. I wonder which 1.8 will get cancelled this year?
AAMAS '09 Workshops
W1 AOSE Agent Oriented Software Engineering
W2 ATOP Agent-based Technologies and applications for enterprise interOPerability
W3 ProMAS Programming MAS
W4 ADMI Agents and Data Mining Interaction
W5 SOCASE Service-Oriented Computing: Agents, Semantics, and Engineering
W6 Trust-09 Trust in MAS
W7 ArgMAS Argumentation in MAS
W8 MABS Multi-Agent Based Simulation
W9 COIN Coordination, Organization, Institutions and Norms
W10 AGS Agents for Games and Simulation
W11 OptMAS Optimization in MAS
W12 MIMS Mixed-Initiative MAS
W13 ALA Adaptive and Learning Agents
W14 WEIN Emergent Intelligence of Networked Agents
W15 EduMAS Educational use for MAS
W16 Standards of Multimodal Dialogue Acts for Embodied Agents
W17 ABSHLE Agent Based Systems for Human Learning and Entertainment
W18 DALT Declarative Agent Languages and Technologies
W19 MSDM Multi-Agent Sequential Decision Making in Uncertain Domains
W20 Empathic Agents
W21 MMAS Massively Multi-Agent Systems: Models, Methods and Tools
W22 BASeWEB Business Agents and the Semantic Web
W23 ATSN Agent Technology for Sensor Networks
W24 AP2PC Agents in P2P computing
W25 ACAN Agent-based Complex Automated Negotiations
W26 ADAPT Agent Design: Adapting from Practice to Theory
W27 AMEC Agent-Mediated Electronic Commerce
http://www.conferences.hu/AAMAS2009/workshops.html
W2 ATOP Agent-based Technologies and applications for enterprise interOPerability
W3 ProMAS Programming MAS
W4 ADMI Agents and Data Mining Interaction
W5 SOCASE Service-Oriented Computing: Agents, Semantics, and Engineering
W6 Trust-09 Trust in MAS
W7 ArgMAS Argumentation in MAS
W8 MABS Multi-Agent Based Simulation
W9 COIN Coordination, Organization, Institutions and Norms
W10 AGS Agents for Games and Simulation
W11 OptMAS Optimization in MAS
W12 MIMS Mixed-Initiative MAS
W13 ALA Adaptive and Learning Agents
W14 WEIN Emergent Intelligence of Networked Agents
W15 EduMAS Educational use for MAS
W16 Standards of Multimodal Dialogue Acts for Embodied Agents
W17 ABSHLE Agent Based Systems for Human Learning and Entertainment
W18 DALT Declarative Agent Languages and Technologies
W19 MSDM Multi-Agent Sequential Decision Making in Uncertain Domains
W20 Empathic Agents
W21 MMAS Massively Multi-Agent Systems: Models, Methods and Tools
W22 BASeWEB Business Agents and the Semantic Web
W23 ATSN Agent Technology for Sensor Networks
W24 AP2PC Agents in P2P computing
W25 ACAN Agent-based Complex Automated Negotiations
W26 ADAPT Agent Design: Adapting from Practice to Theory
W27 AMEC Agent-Mediated Electronic Commerce
http://www.conferences.hu/AAMAS2009/workshops.html
Saturday, January 17, 2009
Most cited MAS articles on google scholar
Here are the top five most cited articles returned by google scholar on a search for "Multi agent":
Ferber, "Multi-Agent Systems: An Introduction to Distributed Artificial Intelligence", 1999
Lux and Marchesi, "SCALING AND CRITICALITY IN A STOCHASTIC MULTI-AGENT MODEL OF A FINANCIAL MARKET", Nature: International weekly journal of science, 1999
Littman, "Markov games as a framework for multi-agent reinforcement learning", ICML, 1994
Nwana, Ndumu, Lee, and Collis, "ZEUS: A toolkit for building distributed multi-agent
systems'' Applied Artificial Intelligence, 1999
Ferber and Gutknecht, “"A meta-model for the analysis and design of organizations in multiagent
systems",” (ICMAS-98).
I guess the golden age for multi-agents was about 10 years ago ....
Ferber, "Multi-Agent Systems: An Introduction to Distributed Artificial Intelligence", 1999
Lux and Marchesi, "SCALING AND CRITICALITY IN A STOCHASTIC MULTI-AGENT MODEL OF A FINANCIAL MARKET", Nature: International weekly journal of science, 1999
Littman, "Markov games as a framework for multi-agent reinforcement learning", ICML, 1994
Nwana, Ndumu, Lee, and Collis, "ZEUS: A toolkit for building distributed multi-agent
systems'' Applied Artificial Intelligence, 1999
Ferber and Gutknecht, “"A meta-model for the analysis and design of organizations in multiagent
systems",” (ICMAS-98).
I guess the golden age for multi-agents was about 10 years ago ....
Friday, January 16, 2009
Multi-Robot Systems for Warehouse Management
From Sciam:
Kiva's robots resemble ground-hugging iRobot Roombas more than the humanoid robot Sonny from the 2004 movie I, Robot. Kiva robots and Roombas, however, are the reality of artificially intelligent robotics. They may not be able to run, jump or speak, but they can efficiently move shelves laden with heavy inventory and clean up messes. And with customers like Staples and Walgreens populating their distribution centers with Kiva's creations, "we're finally seeing massive, multirobot systems at the commercial level," [Peter] Wurman says. "As the scale becomes bigger, the decision-making skills become more important."
http://www.sciam.com/article.cfm?id=artificial-intelligence-robots-rule
Kiva's robots resemble ground-hugging iRobot Roombas more than the humanoid robot Sonny from the 2004 movie I, Robot. Kiva robots and Roombas, however, are the reality of artificially intelligent robotics. They may not be able to run, jump or speak, but they can efficiently move shelves laden with heavy inventory and clean up messes. And with customers like Staples and Walgreens populating their distribution centers with Kiva's creations, "we're finally seeing massive, multirobot systems at the commercial level," [Peter] Wurman says. "As the scale becomes bigger, the decision-making skills become more important."
http://www.sciam.com/article.cfm?id=artificial-intelligence-robots-rule
AAMAS Doctoral Mentoring
Final Call for Doctoral Mentoring Program
The Eighth International Joint Conference on
AUTONOMOUS AGENTS AND MULTIAGENT SYSTEMS (AAMAS 2009)
Budapest, Hungary
Symposium Date: May 10, 2009
Conference Dates: May 10-15, 2009
http://www.conferences.hu/AAMAS2009/
AAMAS 2009 will include a doctoral mentoring program, intended for PhD
students in advanced stages of their research. This program will
provide an opportunity for students to interact closely with
established researchers in their fields, to receive feedback on their
work and to get advice on managing their careers.
Specifically, the goals of the program are:
* To match each student with an established researcher in the
community (who will act as a mentor). The mentor will interact closely
with the student, will provide feedback on research, help form new
contacts, etc.
* To allow students an opportunity to present their work to a
friendly audience of other students as well as mentors.
* To provide students with contacts and professional networking opportunities.
* The doctoral mentoring program will consist of opportunities for
interactions between mentors and their mentorees prior to the
conference, as well as a one day doctoral symposium.
A.1 Submission Requirements
We encourage submissions from PhD students at advanced stages of their
research within the Autonomous Agents and Multi Agent Systems field.
Based on the submissions, the organizing committee will select a group
of students that will be invited to participate in the program.
Participants will be expected to take active part in all doctoral
mentoring program activities.
Each submission should include a set of documents from the students,
and a recommendation letter from the advisor. The submission package
should include:
1. A two-page extended abstract of the student's thesis (in the AAMAS
submission format)
2. A personal research statement (one page)
3. A short (2-page) resume (CV)
4. A recommendation letter from the advisor.
Submissions of the first three items should be sent in electronic form
(PDF), to the doctoral mentoring chair (Stacy Marsella,
mentoring@aamas09.org), by January 31, 2009.
In addition, the dissertation advisor should send the last item (a
letter of recommendation) by e-mail to mentoring@aamas09.org. The
letter should address the expected benefit of attending (to the
student), the significance of the research, and the expected date for
thesis submission. This letter can be sent in either plain text or PDF
format. The student's name must be clearly pointed out in the letter.
A.2 Important Dates
* January 31: Submission package due
* February 28: Acceptance notifications
* March 12: Camera-ready copy due
* May 10: Doctoral Mentoring Symposium
The one-day symposium will be held on the first day of the conference.
Doctoral Mentoring Chair
Stacy Marsella
University of Southern California
marsella [at] ict.usc.edu
Web page:
http://www.conferences.hu/AAMAS2009/mentoring.html
The Eighth International Joint Conference on
AUTONOMOUS AGENTS AND MULTIAGENT SYSTEMS (AAMAS 2009)
Budapest, Hungary
Symposium Date: May 10, 2009
Conference Dates: May 10-15, 2009
http://www.conferences.hu/AAMAS2009/
AAMAS 2009 will include a doctoral mentoring program, intended for PhD
students in advanced stages of their research. This program will
provide an opportunity for students to interact closely with
established researchers in their fields, to receive feedback on their
work and to get advice on managing their careers.
Specifically, the goals of the program are:
* To match each student with an established researcher in the
community (who will act as a mentor). The mentor will interact closely
with the student, will provide feedback on research, help form new
contacts, etc.
* To allow students an opportunity to present their work to a
friendly audience of other students as well as mentors.
* To provide students with contacts and professional networking opportunities.
* The doctoral mentoring program will consist of opportunities for
interactions between mentors and their mentorees prior to the
conference, as well as a one day doctoral symposium.
A.1 Submission Requirements
We encourage submissions from PhD students at advanced stages of their
research within the Autonomous Agents and Multi Agent Systems field.
Based on the submissions, the organizing committee will select a group
of students that will be invited to participate in the program.
Participants will be expected to take active part in all doctoral
mentoring program activities.
Each submission should include a set of documents from the students,
and a recommendation letter from the advisor. The submission package
should include:
1. A two-page extended abstract of the student's thesis (in the AAMAS
submission format)
2. A personal research statement (one page)
3. A short (2-page) resume (CV)
4. A recommendation letter from the advisor.
Submissions of the first three items should be sent in electronic form
(PDF), to the doctoral mentoring chair (Stacy Marsella,
mentoring@aamas09.org), by January 31, 2009.
In addition, the dissertation advisor should send the last item (a
letter of recommendation) by e-mail to mentoring@aamas09.org. The
letter should address the expected benefit of attending (to the
student), the significance of the research, and the expected date for
thesis submission. This letter can be sent in either plain text or PDF
format. The student's name must be clearly pointed out in the letter.
A.2 Important Dates
* January 31: Submission package due
* February 28: Acceptance notifications
* March 12: Camera-ready copy due
* May 10: Doctoral Mentoring Symposium
The one-day symposium will be held on the first day of the conference.
Doctoral Mentoring Chair
Stacy Marsella
University of Southern California
marsella [at] ict.usc.edu
Web page:
http://www.conferences.hu/AAMAS2009/mentoring.html
Welcome
The plan for this blog is to keep readers up-to-date with the world of multi-agent and multi-robot research.
Subscribe to:
Posts (Atom)