Abstract
The price for traveling in a network paid by a single traveler often depends
on decisions made by other passengers. In particular, the time spent on
each link in the journey (e.g. a road or an airport) is known to increase
steeply with the total number of users in real transportation
networks. Different policies to assign passengers to
routes through the network can affect the distribution of traffic and, hence,
the delays experienced by individuals. It is known from game theory that
users can cause unnecessarily long travel times to others and sometimes even
themselves if allowed to individually seek the quickest
paths. However, it remains unknown how inefficient
such decisions based on everybody's self-interest are in real networks. Here
we show that travel times in Boston's road network can be expected to be up
to 30% longer than the ``Social Optimum''. To gain a better understanding
under what circumstances users have to pay such a substantial ``price of
anarchy'', we analyze different models for traffic networks suggested in the
recent literature. A famous result in traffic optimization (known as Braess'
paradox) states that removing links can sometimes reduce effective travel
times. We find that this phenomenon exists in the Boston
network, but improvements are too small to make link deletions a promising
way for controling selfish behavior.
Users
Please
log in to take part in the discussion (add own reviews or comments).