You are on page 1of 19

Multiagent Systems

Course Overview and


© Manfred Huber 2018 1

Course Overview
 Course Description:
Multiagent systems has emerged as an important research area with
applications in many fields of computer science, including artificial
intelligence, e-commerce, sensor networks, distributed computing and
information retrieval, information security, and robotics. In multiagent
systems, multiple autonomous entities with their own objectives have
to interact and make decisions. This course explores techniques for
the modeling, design, decision making, and communication in these
systems. While the course will focus on frameworks, methodologies,
and algorithms, it will investigate (and illustrate) them in the context
of a wide range of application areas, including multi-robot systems,
distributed scheduling and resource allocation, sensor networks,
distributed information extraction, and network security.

© Manfred Huber 2018 2

Course Overview
 Course Topics:
 Representations and modeling
 Game theory:
 Matrix and repeat games, stochastic and Bayesian games
 Auction mechanisms
 Sealed bid and Vickrey auctions, English and Dutch
auctions, combinatorial auctions
 Multiagent Communication
 Multiagent Learning
 Coalitional Game Theory
© Manfred Huber 2018 3
Course Overview
 Prerequisites:
Many of the techniques covered in this course are based on
probabilities and random processes and a basic background in
statistics is required for the course (CSE 5301 or equivalent).
In addition, experience with Algorithms (CSE 5311), Artificial
Intelligence (CSE 5360), and programming will be useful to
perform assignments and projects

© Manfred Huber 2018 4

Course Overview
 Course Page and Materials
Y. Shoham, K. Leyton-Brown, Multiagent Systems: Algorithmic, Game-
Theoretic, and Logical Foundations, Cambridge Press, 2009.
(Available at )

Course web page:

Tentative Office Hours

M 2:00-2:45, W 7:00-8:00, Th 2:00-3:00 ERB 128 or ERB 522
© Manfred Huber 2018 5
Course Overview
 Course Work:
 In-class presentation of a technical paper
 Two homework assignments:
 Two small projects
 Final course project
 Grading Policy:
Presentation \& Class Participation 15 %
Assignments 30 %
Projects 30 %
Final Project 25 %
© Manfred Huber 2018 6
Multiagent Systems and
Distributed Decision Making
 Multiagent Systems:
A system consisting of multiple agents that interact (directly or
indirectly through the environment) and reason and make
decisions individually (generally with incomplete local

 Centralized Systems:
A central coordinator determines the actions that each agent in the
system should take

 Distributed Systems:
Each agent has to determine the action to be taken (including the
exchange of information) based on its local information

© Manfred Huber 2018 7

Collaborative and Competitive
 Collaborative Multiagent Systems:
 Agents have the same desires
 Well defined optimality
 Issues:
 Coordination between distributed agents
 Communication and bandwidth
 Competitive Multiagent Systems:
 Different desires for different agents
 Optimality only defined for individual agents
 Issues:
 Optimal decision making
 Interpretation of communication (agents can lie)
© Manfred Huber 2018 8
Multiagent Decision Making
 Agents and Rationality
To make decisions, agents have to be able to
determine what action is the best for them.
 Rationality:

 Rational agents make the decisions that result in the

highest payoff for them (self-interest)
 Rational agents do not take actions to harm others
 Payoff is quantified in terms of utility

 Multiagent Systems and Optimality

Maximizing an agent’s utility is not always rational
 The Commons problem

© Manfred Huber 2018 9

Multiagent Decision Making
 Multiagent Decisions:
 In competitive systems (even deterministic ones)
optimal decisions often have to be
 An agent’s utility achieved depends not only on
its own actions but also on the actions of the
other agents
 Decision Theory:
 Combines probability, utility theory and rationality
to allow an agent to determine the best action in a
given situation

© Manfred Huber 2018 10

Multiagent Systems

Background - Probability

© Manfred Huber 2018 11

 Bayesian probabilities summarize the effects
of uncertainty on the state of knowledge
 Probabilities represent the values of statistics
 P(o) = (# of times of outcome o) / (# of outcomes)
 All types of uncertainty are incorporated into a
single number
P(H | E)
 Probabilities follow a set of strict axioms

© Manfred Huber 2018 12

 Random variables define the entities of
probability theory
 Propositional random variables:
 E.g.: IsRed, Earthquake
 Multivalued random variables:
 E.g.: Color, Weather
 Potentially Real-Valued
 E.g.: Height, Weight
© Manfred Huber 2018 13
Axioms of Probability
 Probability follows a fixed set of rules
 Propositional random variables:
 P(A)  [0..1]
 P(T) = 1 , P(F) = 0
 P(AB) = P(A) + P(B) – P(AB)
 P(AB) = P(A) P(B|A)
 xValues(X) P(X=x) = 1

© Manfred Huber 2018 14

Probability Syntax
 Unconditional or prior probabilities represent
the state of knowledge before new
observations or evidence
 P(H)
 A probability distribution gives values for all
possible assignments to a random variable
 A joint probability distribution gives values
for all possible assignments to all random
© Manfred Huber 2018 15
Conditional Probability
 Conditional probabilities represent the
probability after certain observations or
facts have been considered
 P(H|E) is the posterior probability of H
after evidence E is taken into account
 Bayes rule allows to derive posterior
probabilities from prior probabilities
 P(H | E) = P(E | H) P(H)/P(E)
© Manfred Huber 2018 16
Conditional Probability
 Probability calculations can be
conditioned by conditioning all terms
 Often it is easier to find conditional
 Conditions can be removed by
 P(H) = E P(H|E) P(E)

© Manfred Huber 2018 17

Joint Distributions
 A joint distribution defines the
probability values for all possible
assignments to all random variables
 Exponential in the number of random
 Conditional probabilities can be computed
from a joint probability distribution
 P(A|B) = P(AB)/P(B)
© Manfred Huber 2018 18
 Inference in probabilistic representation
involves the computation of
(conditional) probabilities from the
available information
 Most frequently the computation of a
posterior probability P(H|E) form a prior
probability P(H) and new evidence E

© Manfred Huber 2018 19