Increasing Traffic Flow through Adaptive Traffic Lights Control

1
Increasing Traffic Flow through Adaptive Traffic
Lights Control

• By
• Michael Mayer
• Donald Bren School of Computer Science,
  University of California, Irvine
• Association for Computing Machinery

2
What is the problem?

• We want to minimize the stopped time globally
• People dont like waiting
• Traffic is a problem that, generally, we all face
  every day
• Minimizing the stopped time will minimize
  gasoline consumption and thus is more green

3
It is NOT a solved problem

• There have been many research papers written on
  this particular problem

• Sensors arent perfect
• Drivers arent perfect
• Lights arent perfect

4
Assumptions made

• Perfect sensors in every intersection
• They are omnificent
• Enough computational power to make quick
  decisions
• We have no guarantee that this is always the case
• However - it appears to be fast enough -
• our simulation runs much faster than real time
• Real implementation appears feasible

5
More reasons why its NOT solved

• Consider P.E.A.S.
• Performance Stopped time
• Environment Fully observable in our
  system
• Actuators The changing traffic lights
• Sensors Unrealistically perfect

6
Infrastructure is expensive

• A normal magnetic coil sensor costs as much as
  2000 each
• Sensors that might be adequate for a system that
  is nearly omnificent cost as much as 26,000 each
  and you would need four per intersection

Source http//www.tfhrc.gov/its/pubs/06108/01.htm
7
How did we approach this problem?

• Taking the P.E.A.S brought up earlier we applied
  artificial intelligence to analyze and reduce the
  drivers waiting time
• Details by Johnny Huynh
• Any questions so far?