Download Speaker Biography: Dr. Talal Rahwan graduated from the

Speaker Biography:
Dr. Talal Rahwan graduated from the Department of Information Technology at the
University of Aleppo in Syria, where he has been awarded Al-Basel Presidential Award for
Academic Achievement (presented annually to students ranked 1st in each department at
Syrian Universities). He was then awarded an Overseas Research Scholarship to do his Ph.D.
at the School of Electronics and Computer Science at the University of Southampton (which
is ranked 2nd in the UK for the Quality of its research). Dr. Rahwan received the British
Computer Society’s Distinguished Dissertation Award, which is annually presented to the
best British Ph.D. in computer science. Since 2006, he has been conducting post doctoral
research at the IAM research group (the biggest Multi-Agent Systems group worldwide) at
the University of Southampton. His research has been published at prestigious Artificial
Intelligence conferences and journals, including AAAI, IJCAI, ECAI, AAMAS, ACM-EC,
JAIR and AIJ.
„Projekt współfinansowany przez Unię Europejską w ramach Europejskiego Funduszu Społecznego”
Coalition Formation in Multi-Agent Systems
By Dr. Talal Rahwan
Open, distributed, computing applications are becoming increasingly common-place in our
society as networked systems and connectivity become the norm. In many cases, these
applications are composed of multiple independent and autonomous actors, known as agents,
which are situated within the same environment. Typically, these agents need to interact with
one another in order to fulfill their objectives or improve their performance. This need comes
from the inevitable interdependencies that exist between the agents’ environment and/or
objectives. In fact, it is the agents’ ability to interact with one another, and compensate for
each others’ deficiencies, that makes multi-agent systems applicable to such a wide variety of
applications.
One of the most fundamental types of interaction is that of Coalition Formation (the topic of
this course). This involves the creation of agile, autonomous, groupings that are goal-directed
and short-lived; i.e. they are formed with a purpose in mind, and are dissolved when that
purpose no longer exists, or when they cease to suit their designed purpose. The area of
coalition formation has received considerable attention in recent research, and has been shown
useful in a wide variety of applications. Examples include multi-sensor networks, distributed
vehicle routing, signal processing, information gathering, and electronic commerce.
This course focuses on the algorithmic aspects of coalition formation. It highlights the main
challenges, and the state-of-the-art search algorithms that have been developed to overcome
these challenges.
Course plan:
Lecture 1: Introduction
 Introduction to Multi-Agent Systems
 The three main activities of Coalition Formation
 Major challenges & Desired solution properties
Lecture 2: Coalition Value Calculation
 Alternative ways of calculating the coalition value
 Alternative approaches to distribute the search space
Lecture 3: Caolition Structure Generation in Characteristic function Games (Part 1)
 Analyzing the space of possible coalition structures
 Can we search only part of the space and still get guarantees?
„Projekt współfinansowany przez Unię Europejską w ramach Europejskiego Funduszu Społecznego”
Lecture 4: Caolition Structure Generation in Characteristic Function Games (Part 2)
 What are Characteristic Function Games (CFGs)?
 Analyzing the search space in CFGs
Lecture 5: Coalition Structure Generation in Characteristic Function Games (Part 3)
 Search Technique 1: Branch- and-Bound
 Search Technique 2: Dynamic Programming
Lecture 6: Coalition Structure Generation in Partition Function Games (Part 1)
 What are Partition Function Games (PFGs)?
 Analyzing the search space in PFGs
 Computer bounds on coalition values
Lecture 7: Coalition Structure Generation in Partition Function Games (Part 2)
 Can we search only part of the search space and still get guarantees?
 Methods for speeding up the search
Lecture 7: Constraint Satisfaction in Coalition Formation
 What constraints can we place on the search space?
 Using the constraints to prune the search space
Lecture 8: Distributed Coalition Formation
 How to search for the optimal coalition structure in a distributed manner.
Lecture 9: Coalition Formation in multi-sensor networks
 How can sensors coordinate their activities to improve the overall performance
 Analyzing the search space of the proposed system
 Solving this problem using some of the techniques that were presented during the
course
Lecture 10: Summary & possible extensions
 Possible extensions to the topics that were discussed throughout the course
„Projekt współfinansowany przez Unię Europejską w ramach Europejskiego Funduszu Społecznego”