University of California Transportation Center

The city evacuation problem is analyzed physically at the freeway and network levels. On a freeway, a macroscopic approach is used to identify the critical bottlenecks that determine the system’s evacuation capacity. Knowledge of these bottlenecks leads to the development of an input control strategy that maximizes exit flows at all times, effectively minimizing total evacuation time. The optimality results are true for the complete system and for “population nests”. The strategy, called innermost first out (InFO), has many other benefits: it is decentralized, adaptive and robust. Additionally, since the strategy gives priority to upstream, most-at-risk residents, InFO is likely to be socially acceptable. Finally, relaxed versions of the strategy exist, giving flexibility to freeway evacuation management.

At the network level, a tree-shaped topology allows for similar results to be obtained. Specifically, a tree-based innermost first out (T-InFO) strategy is developed, combining InFO with an intuitive routing scheme. It is shown that if a reasonable driver adaptive behavior can be assumed for the local access streets, then T-InFO maximizes exit flows and minimizes evacuation time for population nests and therefore the complete system. Similar to InFO, T-InFO has the following benefits when implemented in a tree-shaped network: decentralization, adaptiveness, robustness, and social optimality. Due to these reasons, the strategies proposed in this dissertation have the potential to greatly improve current traffic management practices in emergency evacuations.