| Abstracts of talks by the Speakers of DAMAS |
| |
|
| Agent Organisation & Communication |
|
| Speaker: Dr Frank Dignum |
|
| From: Utrecht University, Netherlands |
|
| |
| Just programming a number of agents and putting them together does not make a good multi-agent system yet. The agents in the system should work together, such that the system as a whole will reach some predetermined goals. The situation gets even more complicated when agents are coming from different parties and have to work together. In this class we will discuss the concept of agent organizations. Which are the models that need to be defined, how do the agents fit into an organization model and how is it implemented?
In such an organization communication is of the utmost importance for reaching the goal of the system. In the second part of the class we will discuss the issues surrounding agent communication both from a theoretical as well as a practical perspective.
|
| |
| |
|
| TAC Hands on Session |
|
| Speaker: Dr Alex Rogers |
|
| From: University of Southampton, UK |
|
| |
|
The annual Trading Agent Competitions (TAC) provide a competitive benchmarking environment for agent trading strategies, and in this hands on lab session, we look in detail at the original TAC Travel game. In particular, we’ll discuss the various markets in which the agents must trade and the challenges that these present, we’ll download and run the agentware required to compete in the game, and we’ll see how simple trading strategies are actually implemented in Java. Finally, after modifying these strategies, we’ll run our own tournament.
|
|
| |
| |
|
| Trading Agent Design and Analysis (Agent Trading Competition – old title) |
|
| Speaker: Prof Michael Wellman |
|
| From: University of Michigan, USA |
|
| |
|
This tutorial frames and motivates the problem of developing automated
trading strategies for electronic markets. E-commerce increasingly
makes use of autonomous bidding agents, computer programs that bid in electronic markets without direct human intervention. Automated
bidding strategies for an auction of a single good with a known
valuation are fairly straightforward to design; designing strategies
for simultaneous auctions with interdependent valuations is a more complex undertaking. This tutorial presents algorithmic advances and bidding agent
architectures that have emerged from recent work in this fast-growing
area of research in academia and industry. It surveys the
state-of-the-art in analyzing strategies for basic market games,
covers examples of more complex (intractable) market scenarios, and
presents a general methodology (empirical and game-theoretic) for
trading agent design and analysis.
|
|
| |
| |
|
| Mechanism Design for MAS |
|
| Speaker: Dr Vincent Conitzer |
|
| From: Duke University, USA |
|
| |
|
In multiagent systems, the agents often have to make a joint decision even
though they have different preferences over the alternatives. For example,
the agents may have to decide on a joint plan, or on an allocation of tasks
or resources among themselves. In general, each agent initially knows only
its own preferences over the alternatives; the agents then report their
preferences in some form, and based on these reports, one of the
alternatives is chosen. Auctions and elections can both be viewed as
special cases of this setup. A key issue is that an agent may lie about
its preferences if it perceives this to be in its interest. The challenge
is to design the outcome-choosing mechanism in such a way that a good
result is obtained nevertheless -- for example, by making it optimal for
each agent to report its true preferences. Mechanism design has
traditionally been studied by game theorists, economists, and political
scientists, but it is increasingly studied by computer scientists.
The first part of this tutorial will cover the basics of "classical"
mechanism design, including basic definitions, the revelation principle,
Vickrey-Clarke-Groves mechanisms, and impossibility results. The second
part will cover computational aspects of mechanism design, including
whether and how mechanisms' outcomes can be efficiently computed;
algorithms for automatically designing the entire mechanism; and
limitations of classical results in the face of computationally bounded
agents.
Example applications will be provided throughout.
|
|
| |
| |
|
| Robot and Human-Robot Teams |
|
| Speaker: Dr M Bernardine Dias |
|
| From: Carnegie Mellon University, USA |
|
| |
|
Robots are playing an increasingly important role in the world today and multi-robot systems are now becoming a popular solution for a variety of complex tasks such as mapping, exploration, and assembly. Moreover, as these robots become an integral part of society, research to enable effective human-robot teams has been growing in importance. Regardless of the composition of the team, an important factor for effective team performance is coordination, which involves the allocation and collective execution of sub-tasks through an efficient mechanism. This session will provide an overview of various methods for coordinating robot teams and human-robot teams and discuss their advantages and disadvantages. The tutorial covers theoretical, experimental, and implementation factors and makes no assumptions about the background of the audience. Hence, the session will begin with an introduction to robotics, followed by an introduction to multi-robot systems and human-robot teams, will review a variety of the relevant approaches and methodologies, and will especially detail market-based methods applied to coordinating these teams.
|
|
| |
| |
|
| Introduction to MAS |
|
| Speaker: Dr I. Rahwan & Dr S. Abdallah |
|
| From: The British University in Dubai, UAE |
|
| |
|
The abstract will be added soon.
|
|
| |
| |
|
| Learning in MAS |
|
| Speaker: Dr Michael Rovatsos |
|
| From: University of Edinburgh, UK |
|
| |
|
In open multiagent environments composed of heterogeneous agents which
may pertain to different human users and may be pursuing conflicting
design objectives, it is often impossible to establish appropriate
methods for coordinating the behaviours of different agents at design
time. This is because the internal design of agents may be opaque
for their peers and hence the concrete interactions that the agents will
have cannot be anticipated. Learning and adaptation are therefore of crucial
importance to the design of flexible, intelligent agents that have the
ability to achieve their own design objectives in such inherently
dynamic, decentralised environments. Today, the study of multiagent
learning methods assumes a prominent role in multiagent systems
research, and has produced a body of very diverse techniques.
This tutorial will provide an introduction to the field. In the first
part, we will review basic notions of machine learning which provide
the basis for algorithms employed in adaptive multiagent systems, and
we will survey a number of well-known approaches from the literature
concerned with learning with, from, and about other agents. The second
part will be devoted to a more in-depth survey of multiagent
reinforcement learning algorithms, since this area has emerged as the
currently most active sub-area of the field of multiagent learning and
is useful to highlight fundamental problems in multiagent learning
(which are still unsolved) and discuss potential solutions.
|
|
| |
| |
|
| Argumentation in MAS |
|
| Speaker: Dr Chris Reed |
|
| From: University of Dundee, UK |
|
| |
|
The theory of argument stretches back to classical times, and is still an
active area of philosophical research. Since at least the early 1980s, argument
has inspired theory and practice in artificial intelligence, and by the mid
1990s, the road between argumentation theory and AI models of argument was
open to increasing traffic. This lecture introduces the background to
argumentation and gives a very brief overview of modern argumentation theory.
It then analyses the two areas in which theories of argument have had greatest
impact: nonmonotonic reasoning and inter-agent communication. The target is to
bring students up to date with key results from research in the ArgMAS and
related communities from the past decade. Throughout, we will focus on the
motivation for each major theoretical approach in order to offer insight
into the reasons for the current and growing popularity both of individual
techniques and of argumentation as a whole in multi-agent systems research.
|
|
| |
| |
