Methods for analyzing routing games: Information design, risk-averseness, and Braess's paradox

We study ways in which information about traffic networks can be used to achieve social objectives, such as decreasing experienced travel time.
First we study how behaviour of drivers changes when the total amount of traffic changes. We give an exact characterization of this change in behaviour, which is computationally feasible to obtain. We then use insights obtained from deriving this characterization to consider the difficult problem of detecting Braess’s paradox in a network, where removal of a road leads to decreased travel time for all drivers. We give some new, efficient methods for detecting this phenomenon, and also show that in some cases the existence of Braess’s paradox in a network may be a good thing.
Next we study traffic networks with potentially unpredictable travel costs. We investigate a scenario a central planner can strategically withhold information from drivers on the road to prevent congestion and benefit all road users. However, the planner’s strategy depends upon the prior beliefs about the roads that the drivers adhere to, and we study how a planner can derive these beliefs by observing the behaviour of the drivers.
Finally we study the scenario where drivers are risk-averse, and can thus avoid roads that are quick on average, but can be significantly slowed in some cases.