UTCS Colloquium/AI: Shimon Whiteson/University of Amsterdam: "Multi-Agent Reinforcement Learning for Urban Traffic Control using Coordination Graphs " TAY 3.128, Monday, June 22, 2009 11:00 a.m.

Contact Name: 
Jenna Whitney
Jun 22, 2009 11:00am - 12:00pm

There is a signup schedule for this event (UT EID required).

Type o

f Talk:  UTCS Colloquium/AI

Speaker/Affiliation:  Shimon

Whiteson/University of Amsterdam

Date/Time:  Monday, June 22,
2009 11:00 a.m.

Location:  TAY 3.128

Host:  Peter


Talk Title: "Multi-Agent Reinforcement Learning for Urban
Traffic Control using Coordination Graphs"

Talk Abstract:


p>Since traffic jams are ubiquitous in the modern world, optimizing the be

havior of traffic lights for efficient traffic flow is a critically importa

nt goal. Though most current traffic lights use simple heuristic protocols

, more efficient controllers can be discovered automatically via multi-agen

t reinforcement learning, where each agent controls a single traffic light

. However, in previous work on this approach, agents select only locally

optimal actions without coordinating their behavior. In this talk I describ

e how to extend this approach to include explicit coordination between neig

hboring traffic lights. Coordination is achieved using the max-plus algorit

hm, which estimates the optimal joint action by sending locally optimized

messages among connected agents. I present the first application of max-plu

s to a large-scale problem and provide empirical evidence that max-plus per

forms well on cyclic graphs, though it has been proven to converge only fo

r tree-structured graphs. I also discuss our future plans for tackling traf

fic control problems by formalizing them as loosely-coupled multi-objective
multiagent control problems.

Speaker Bio:

Shimon Whiteson is a

n Assistant Professor in the Intelligent Autonomous Systems group at the Un

iversity of Amsterdam. He received his PhD in 2007 from the University of T

exas at Austin, working with Peter Stone. His current research interests f

ocus on single- and multi-agent decision-theoretic planning and reinforceme

nt learning.